There appears to be a widespread belief that some processes of selecting cryptosystems are less risky than other processes. As a case study of quantifying the difference in risks, this paper compares the currently-known-failure rates of three large groups of cryptosystems: (1) the round-1 submissions to the NIST Post-Quantum Cryptography Standardization Project, (2) the round-1 submissions not broken by the end of round 1, and (3) the round-1 submissions selected by NIST for round 2 of the same project. These groups of cryptosystems turn out to have currently-known-failure rates that are strikingly high, and that include statistically significant differences across the groups, not matching the pattern of differences that one might expect. Readers are cautioned that the actual failure rates could be much higher than the currently-known-failure rates.

The Dual-Sieve Attack on Learning with Errors (LWE), or more generally Bounded Distance Decoding (BDD), has seen many improvements in the recent years, and ultimately led to claims that it outperforms the primal attack against certain lattice-based schemes in the PQC standardization process organised by NIST. However, the work of Ducas--Pulles (Crypto '23) revealed that the so-called "Independence Heuristic", which all recent dual attacks used, leads to wrong predictions in a contradictory regime, which is relevant for the security of cryptoschemes. More specifically, the stated distributions of scores for the actual solution and for incorrect candidates were both incorrect. In this work, we propose to use the weaker heuristic that the output vectors of a lattice sieve are uniformly distributed in a ball. Under this heuristic, we give an analysis of the score distribution in the case of an error of fixed length. Integrating over this length, we extend this analysis to any radially distributed error, in particular the gaussian as a fix for the score distribution of the actual solution. This approach also provides a prediction for the score of incorrect candidates, using a ball as an approximation of the Voronoi cell of a lattice. We compare the predicted score distributions to extensive experiments, and observe them to be qualitatively and quantitatively quite accurate. This constitutes a first step towards fixing the analysis of the dual-sieve attack: we can now accurately estimate false-positives and false-negatives. Now that the analysis is fixed, one may consider how to fix the attack itself, namely exploring the opportunities to mitigate a large number of false-positives.

Credential tweaking attacks use breached passwords to generate semantically similar passwords and gain access to victims' services. These attacks sidestep the first generation of compromised credential checking (C3) services. The second generation of compromised credential checking services, called "Might I Get Pwned" (MIGP), is a privacy-preserving protocol that defends against credential tweaking attacks by allowing clients to query whether a password or a semantically similar variation is present in the server's compromised credentials dataset. The desired privacy requirements include not revealing the user's entered password to the server and ensuring that no compromised credentials are disclosed to the client. In this work, we formalize the cryptographic leakage of the MIGP protocol and perform a security analysis to assess its impact on the credentials held by the server. We focus on how this leakage aids breach extraction attacks, where an honest-but-curious client interacts with the server to extract information about the stored credentials. Furthermore, we discover additional leakage that arises from the implementation of Cloudflare's deployment of MIGP. We evaluate how the discovered leakage affects the guessing capability of an attacker in relation to breach extraction attacks. Finally, we propose MIGP 2.0, a new iteration of the MIGP protocol designed to minimize data leakage and prevent the introduced attacks.

Decentralized Finance (DeFi) is a new paradigm in the creation, distribution, and utilization of financial services via the integration of blockchain technology. Our research conducts a comprehensive introduction and meticulous classification of various DeFi applications. Beyond that, we thoroughly analyze these risks from both technical and economic perspectives, spanning multiple layers. We point out research gaps and revenues, covering technical advancements, innovative economics, and sociology and ecology optimization.

This study focuses on spotting and stopping new types of online threats by improving the UGRansome dataset to detect unusual activity in real-time. By blending different machine learning methods, like naïve tree-based ensemble learning and recursive feature elimination (RFE), the research achieves a high accuracy rate of 97%. Naïve Bayes (NB) stands out as the most effective classifier. The suggested setup, combining gradient boosting (GB) and random forest (RF) with NB, effectively identifies and prevents unknown vulnerabilities in computer systems. UGRansome successfully blocks over 100 kilobits per second (kbps) of harmful online traffic by using details pinpointed by the RFE method, specifically uniform resource locators (URLs). This outperforms existing Intrusion Detection System (IDS) datasets. It's particularly good at stopping secure shell attacks, proving the dataset's usefulness in making networks safer. This research marks significant progress in detecting intrusions. The NB model excels in accuracy, precision, and remembering patterns, especially in identifying new threats. Moreover, the suggested naïve tree-based ensemble model shows outstanding accuracy, standing out as the best-performing technique among all models studied. Applying the UGRansome properties-based rule noticeably changes how traffic is sorted, decreasing unknown traffic while increasing unclassified traffic, which requires more investigation.

An important criteria to assert the security of a cryptographic primitive is its resistance against differential cryptanalysis. For word-oriented primitives, a common technique to determine the number of rounds required to ensure the immunity against differential distinguishers is to consider truncated differential characteristics and to count the number of active S-boxes. Doing so allows one to provide an upper bound on the probability of the best differential characteristic with a reduced computational cost. However, in order to design very efficient primitives, it might be needed to evaluate the probability more accurately. This is usually done in a second step, during which one tries to instantiate truncated differential characteristics with actual values and computes its corresponding probability. This step is usually done either with ad-hoc algorithms or with CP, SAT or MILP models that are solved by generic solvers. In this paper, we present a generic tool for automatically generating these models to handle all word-oriented ciphers. Furthermore the running times to solve these models are very competitive with all the previous dedicated approaches.

CRYSTALS-Kyber is a key-encapsulation mechanism, whose security is based on the hardness of solving the learning-with-errors (LWE) problem over module lattices. As in its specification, Kyber prescribes the usage of the Number Theoretic Transform (NTT) for efficient polynomial multiplication. Side-channel assisted attacks against Post-Quantum Cryptography (PQC) algorithms like Kyber remain a concern in the ongoing standardization process of quantum-computer-resistant cryptosystems. Among the attacks, correlation power analysis (CPA) is emerging as a popular option because it does not require detailed knowledge about the attacked device and can reveal the secret key even if the recorded power traces are extremely noisy. In this paper, we present a two-step attack to achieve a full-key recovery on lattice-based cryptosystems that utilize NTT for efficient polynomial multiplication. First, we use CPA to recover a portion of the secret key from the power consumption of these polynomial multiplications in the decryption process. Then, using the information, we are able to fully recover the secret key by constructing an LWE problem with a smaller lattice rank and solving it with lattice reduction algorithms. Our attack can be expanded to other cryptosystems using NTT-based polynomial multiplication, including Saber. It can be further parallelized and experiments on simulated traces show that the whole process can be done within 20 minutes on a 16-core machine with 200 traces. Compared to other CPA attacks targeting NTT in the cryptosystems, our attack achieves lower runtime in practice. Furthermore, we can theoretically decrease the number of traces needed by using lattice reduction if the same measurement is used. Our lattice attack also outperforms the state-of-the-art result on integrating side-channel hints into lattices, however, the improvement heavily depends on the implementation of the NTT chosen by the users.

Designing novel symmetric-key primitives for advanced protocols like secure multiparty computation (MPC), fully homomorphic encryption (FHE) and zero-knowledge proof systems (ZK), has been an important research topic in recent years. Many such existing primitives adopt quite different design strategies from conventional block ciphers. Notable features include that many of these ciphers are defined over a large finite field, and that a power map is commonly used to construct the nonlinear component due to its efficiency in these applications as well as its strong resistance against the differential and linear cryptanalysis. In this paper, we target the MPC-friendly ciphers AIM and RAIN used for the post-quantum signature schemes AIMer (CCS 2023 and NIST PQC Round 1 Additional Signatures) and Rainier (CCS 2022), respectively. Specifically, we can find equivalent representations of 2-round RAIN and full-round AIM, respectively, which make them vulnerable to either the polynomial method, or the crossbred algorithm, or the fast exhaustive search attack. Consequently, we can break 2-round RAIN with the 128/192/256-bit key in only $2^{111}/2^{170}/2^{224}$ bit operations. For full-round AIM with the 128/192/256-bit key, we could break them in $2^{136.2}/2^{200.7}/2^{265}$ bit operations, which are equivalent to about $2^{115}/2^{178}/2^{241}$ calls of the underlying primitives. In particular, our analysis indicates that AIM does not reach the required security levels by the NIST competition.

Arithmetization-Oriented primitives are the building block of advanced cryptographic protocols such as Zero-Knowledge proof systems. One approach to designing such primitives is the HADES design strategy which aims to provide an efficient way to instantiate generalizing substitution-permutation networks to include partial S-box rounds. A notable instance of HADES, introduced by Grassi \emph{et al.} at USENIX Security '21, is Poseidon. Because of its impressive efficiency and low arithmetic complexity, Poseidon is a popular choice among the designers of integrity-proof systems. An updated version of Poseidon, namely, Poseidon2 was published at AfricaCrypt '23 aiming to improve the efficiency of Poseidon by optimizing its linear operations. In this work, we show some caveats in the security argument of HADES against algebraic attacks and quantify the complexity of Gr\"{o}bner basis attacks. We show that the complexity of the attack is lower than claimed with the direct implication that there are cases where the recommended number of rounds is insufficient for meeting the claimed security. Concretely, the complexity of a Gr\"{o}bner basis attack for an instance of Poseidon with 1024 bits of security is 731.77 bits and the original security argument starts failing already at the 384-bit security level. Since the security of Poseidon2 is derived from the security of Poseidon, the same analysis applies to the instances of Poseidon2. The results were shared with the designers and the security arguments were updated accordingly.

This paper presents a comprehensive analysis of the verification algorithm of the CRYSTALS-Dilithium, focusing on a C reference implementation. Limited research has been conducted on its susceptibility to fault attacks, despite its critical role in ensuring the scheme’s security. To fill this gap, we investigate three distinct fault models - randomizing faults, zeroizing faults, and skipping faults - to identify vulnerabilities within the verification process. Based on our analysis, we propose a methodology for forging CRYSTALS-Dilithium signatures without knowledge of the secret key. Instead, we leverage specific types of faults during the verification phase and some properties about public parameters to make these signatures accepted. Additionally, we compared different attack scenarios after identifying sensitive operations within the verification algorithm. The most effective requires potentially fewer fault injections than targeting the verification check itself. Finally, we introduce a set of countermeasures designed to thwart all the identified scenarios rendering the verification algorithm intrinsically resistant to the presented attacks.

Deploying cryptography on embedded systems requires security against physical attacks. At CHES 2019, M&M was proposed as a combined countermeasure applying masking against SCAs and information-theoretic MAC tags against FAs. In this paper, we show that one of the protected AES implementations in the M&M paper is vulnerable to a zero-value SIFA2-like attack. A practical attack is demonstrated on an ASIC board. We propose two versions of the attack: the first follows the SIFA approach to inject faults in the last round, while the second one is an extension of SIFA and FTA but applied to the first round with chosen plaintext. The two versions work at the byte level, but the latter version considerably improves the efficiency of the attack. Moreover, we show that this zero-value SIFA2 attack is specific to the AES tower-field decomposed S-box design. Hence, such attacks are applicable to any implementation featuring this AES S-box architecture. Then, we propose a countermeasure that prevents these attacks. We extend M&M with a fine-grained detection-based feature capable of detecting the zero-value glitch attacks. In this effort, we also solve the problem of a combined attack on the ciphertext output check of M&M scheme by using Kronecker's delta function. We deploy the countermeasure on FPGA and verify its security against both fault and side-channel analysis with practical experiments.

Let's consider a scenario where the server encrypts data using AES-CBC without authentication and then sends only the encrypted ciphertext through TLS (without IV). Then, having a padding oracle, we managed to recover the initialization vector and the sensitive data, doing a cybersecurity audit for a Chilean company.

We show that the Nikooghadam-Shahriari-Saeidi authentication and key agreement scheme [J. Inf. Secur. Appl., 76, 103523 (2023)] cannot resist impersonation attack, not as claimed. An adversary can impersonate the RFID reader to cheat the RFID tag. The drawback results from its simple secret key invoking mechanism. We also find it seems difficult to revise the scheme due to the inherent flaw.

Non-invasive fault injection attacks have emerged as significant threats to a spectrum of microelectronic systems ranging from commodity devices to high-end customized processors. Unlike their invasive counterparts, these attacks are more affordable and can exploit system vulnerabilities without altering the hardware physically. Furthermore, certain non-invasive fault injection strategies allow for remote vulnerability exploitation without the requirement of physical proximity. However, existing studies lack extensive investigation into these attacks across diverse target platforms, threat models, emerging attack strategies, assessment frameworks, and mitigation approaches. In this paper, we provide a comprehensive overview of contemporary research on non-invasive fault injection attacks. Our objective is to consolidate and scrutinize the various techniques, methodologies, target systems susceptible to the attacks, and existing mitigation mechanisms advanced by the research community. Besides, we categorize attack strategies based on several aspects, present a detailed comparison among the categories, and highlight research challenges with future direction. By underlining and discussing the landscape of cutting-edge, non-invasive fault injection, we hope more researchers, designers, and security professionals examine the attacks further and take such threats into consideration while developing effective countermeasures.

We analyze REDOG, a public-key encryption system submitted to the Korean competition on post-quantum cryptography. REDOG is based on rank-metric codes. We prove its incorrectness and attack its implementation, providing an efficient message recovery attack. Furthermore, we show that the security of REDOG is much lower than claimed. We then proceed to mitigate these issues and provide two approaches to fix the decryption issue, one of which also leads to better security.

Numerous fault models have been developed, each with distinct characteristics and effects. These models should be evaluated in light of their costs, repeatability, and practicability. Moreover, there must be effective ways to use the injected fault to retrieve the secret key, especially if there are some countermeasures in the implementation. In this paper, we introduce a new fault analysis technique called ``linked fault analysis'' (LFA), which can be viewed as a more powerful version of well-known fault attacks against implementations of symmetric primitives in various circumstances, especially software implementations. For known fault analyses, the bias over the faulty value or the relationship between the correct value and the faulty one, both produced by the fault injection serve as the foundations for the fault model. In the LFA, however, a single fault involves two intermediate values. The faulty target variable, $u'$, is linked to a second variable, $v$, such that a particular relation holds: $u'=l(v)$. We show that LFA lets the attacker perform fault attacks without the input control, with much fewer data than previously introduced fault attacks in the same class. Also, we show two approaches, called LDFA and LIFA, that show how LFA can be utilized in the presence or absence of typical redundant-based countermeasures. Finally, we demonstrate that LFA is still effective, but under specific circumstances, even when masking protections are in place. We performed our attacks against the public implementation of AES in ATMEGA328p to show how LFA works in the real world. The practical results and simulations validate our theoretical models as well.

Physical attacks are serious threats to cryptosystems deployed in the real world. In this work, we propose a microarchitectural end-to-end attack methodology on generic lattice-based post-quantum key encapsulation mechanisms to recover the long-term secret key. Our attack targets a critical component of a Fujisaki-Okamoto transform that is used in the construction of almost all lattice-based key encapsulation mechanisms. We demonstrate our attack model on practical schemes such as Kyber and Saber by using Rowhammer. We show that our attack is highly practical and imposes little preconditions on the attacker to succeed. As an additional contribution, we propose an improved version of the plaintext checking oracle, which is used by almost all physical attack strategies on lattice-based key-encapsulation mechanisms. Our improvement reduces the number of queries to the plaintext checking oracle by as much as 39% for Saber and approximately 23% for Kyber768. This can be of independent interest and can also be used to reduce the complexity of other attacks.

The Permuted Kernel Problem (PKP) asks to find a permutation which maps an input matrix into the kernel of some given vector space. The literature exhibits several works studying its hardness in the case of the input matrix being mono-dimensional (i.e., a vector), while the multi-dimensional case has received much less attention and, de facto, only the case of a binary ambient finite field has been studied. The Subcode Equivalence Problem (SEP), instead, asks to find a permutation so that a given linear code becomes a subcode of another given code. At the best of our knowledge, no algorithm to solve the SEP has ever been proposed. In this paper we study the computational hardness of solving these problems. We first show that, despite going by different names, PKP and SEP are exactly the same problem. Then we consider the state-of-the-art solver for the mono-dimensional PKP (namely, the KMP algorithm, proposed by Koussa, Macario-Rat and Patarin), generalize it to the multi-dimensional case and analyze both the finite and the asymptotic regimes. We further propose a new algorithm, which can be thought of as a refinement of KMP. In the asymptotic regime our algorithm does not improve on KMP but, in the finite regime (and for parameters of practical interest), we achieve significant improvements, especially for the multi-dimensional version of PKP. As an evidence, we show that it is the fastest algorithm to attack several recommended instances of cryptosystems based on PKP. As a side-effect, given the mentioned equivalence between PKP and SEP, all the algorithms we analyze in this paper can be used to solve instances of the latter problem.

We demonstrate that a passive network attacker can opportunistically obtain private RSA host keys from an SSH server that experiences a naturally arising fault during signature computation. In prior work, this was not believed to be possible for the SSH protocol because the signature included information like the shared Diffie-Hellman secret that would not be available to a passive network observer. We show that for the signature parameters commonly in use for SSH, there is an efficient lattice attack to recover the private key in case of a signature fault. We provide a security analysis of the SSH, IKEv1, and IKEv2 protocols in this scenario, and use our attack to discover hundreds of compromised keys in the wild from several independently vulnerable implementations.

Private Set Intersection Cardinality(PSI-CA) is a type of secure two-party computation. It enables two parties, each holding a private set, to jointly compute the cardinality of their intersection without revealing any other private information about their respective sets. In this paper, we manage to break two PSI-CA protocols by recovering the specific intersection items in polynomial time. Among them, the PSI-CA protocol proposed by De Cristofaro et al. in 2012 is the most popular PSI-CA protocol based on the Google Scholar search results and it is still deemed one of the most efficient PSI-CA protocols. In this paper, we also propose several solutions to these protocols' security problems.

The idea of balancing the side-channel leakage in software was proposed more than a decade ago. Just like with other hiding-based countermeasures, the goal is not to hide the leakage completely but to significantly increase the effort required for the attack. Previous approaches focused on two directions: either balancing the Hamming weight of the processed data or deriving the code by using stochastic leakage profiling. In this brief, we build upon these results by proposing a novel approach that combines the two directions. We provide the theory behind our encoding scheme backed by experimental results on a 32-bit ARM Cortex-M4 microcontroller. Our results show that such a combination gives better side-channel resistance properties than each of the two methods separately.

This paper presents full round distinguishing and key recovery attacks on lightweight block cipher SAND-2 with 64-bit block size and 128-bit key size, which appears to be a mixture of the AND-Rotation-XOR (AND-RX) based ciphers SAND and ANT. However, the security arguments against linear and some other attacks are not fully provided. In this paper, we find that the combination of a SAND-like nibble-based round function and ANT-like bit-based permutations will cause dependencies and lead to iterative linear and differential trails with high probabilities. By exploiting these, full round distinguishing attacks on SAND-2 work with $2^{46}$ queries for linear and $2^{58.60}$ queries for differential in the single-key setting. Then, full round key recovery attacks are also mounted, which work with the time complexity $2^{48.23}$ for linear and $2^{64.10}$ for differential. It should be noted that the dependency observed in this paper only works for SAND-2 and will not threaten SAND and ANT. From the point of designers, our attacks show the risk of mixing the parts of different designs, even though each of them is well-studied to be secure.

Differential-Linear (DL) cryptanalysis is a well known cryptanalytic technique that combines differential and linear cryptanalysis. Over the years, multiple techniques were proposed to increase its strength and applicability. Two relatively recent ones are: The partitioning technique by Leurent and the use of neutral bits adapted by Beierle et al. to DL cryptanalysis. In this paper we compare these techniques and discuss the possibility of using them together to achieve the best possible DL attacks. We study the combination of these two techniques and show that in many cases they are indeed compatible. We demonstrate the strength of the combination in two ways. First, we present the first DL attack on 4-round Xoodyak and an extension to 5-round in the related key model. We show that the attacks are possible only by using these two techniques simultaneously. In addition, using the combination of the two techniques we improve a DL attack on 9-round DES. We show that the partitioning technique mainly reduces the time complexity, and the use of neutral bits mainly reduces the data complexity, while the combination of them reduces both the time and data complexities.

We show that the Yang et al.'s key agreement scheme [Future Gener. Comput. Syst., 145, 415-428 (2023)] is flawed. (1) There are some inconsistent computations, which should be corrected. (2) The planned route of a target vehicle is almost exposed. The scheme neglects the basic requirement for bit-wise XOR, and tries to encrypt the route by the operator. The negligence results in some trivial equalities. (3) The scheme is insecure against impersonation attack launched by the next roadside unit.

Learning with Errors (LWE) is a hard math problem underpinning many proposed post-quantum cryptographic (PQC) systems. The only PQC Key Exchange Mechanism (KEM) standardized by NIST is based on module~LWE, and current publicly available PQ Homomorphic Encryption (HE) libraries are based on ring LWE. The security of LWE-based PQ cryptosystems is critical, but certain implementation choices could weaken them. One such choice is sparse binary secrets, desirable for PQ HE schemes for efficiency reasons. Prior work, SALSA, demonstrated a machine learning-based attack on LWE with sparse binary secrets in small dimensions ($n \le 128$) and low Hamming weights ($h \le 4$). However, this attack assumes access to millions of eavesdropped LWE samples and fails at higher Hamming weights or dimensions. We present PICANTE, an enhanced machine learning attack on LWE with sparse binary secrets, which recovers secrets in much larger dimensions (up to $n=350$) and with larger Hamming weights (roughly $n/10$, and up to $h=60$ for $n=350$). We achieve this dramatic improvement via a novel preprocessing step, which allows us to generate training data from a linear number of eavesdropped LWE samples ($4n$) and changes the distribution of the data to improve transformer training. We also improve the secret recovery methods of SALSA and introduce a novel cross-attention recovery mechanism allowing us to read off the secret directly from the trained models. While PICANTE does not threaten NIST's proposed LWE standards, it demonstrates significant improvement over SALSA and could scale further, highlighting the need for future investigation into machine learning attacks on LWE with sparse binary secrets.

Deep learning-based profiling side-channel analysis has gained widespread adoption in academia and industry due to its ability to uncover secrets protected by countermeasures. However, to exploit this capability, an adversary must have access to a clone of the targeted device to obtain profiling measurements and know secret information to label these measurements. Non-profiling attacks avoid these constraints by not relying on secret information for labeled data. Instead, they attempt all key guesses and select the most successful one. Deep learning approaches form the foundation of several non-profiling attacks, but these methods often suffer from high computational complexity and limited performance in practical applications. This work explores the performance of multi-output regression (MOR) models in side-channel analysis. We start with the recently proposed multi-output regression (MOR) approach for non-profiling side-channel analysis. Then, we significantly improve its performance by updating the 1) loss function, 2) distinguisher, and 3) employing a novel concept of validation set to reduce overfitting. We denote our approach as MORE - Multi-Output Regression Enhanced, which emphasizes significantly better attack performance than MOR. Our results demonstrate that combining the MORE methodology, ensembles, and data augmentation presents a potent strategy for enhancing non-profiling side-channel attack performance and improving the reliability of distinguishing key candidates.

Code-based cryptography has received a lot of attention recently because it is considered secure under quantum computing. Among them, the QC-MDPC based scheme is one of the most promising due to its excellent performance. QC-MDPC based scheme is usually subject to a small rate of decryption failure, which can leak information about the secret key. This raises two crucial problems: how to accurately estimate the decryption failure rate and how to use the failure information to recover the secret key. However, the two problems are challenging due to the difficulty of geometrically characterizing the bit-flipping decoder employed in QC-MDPC, such as using decoding radius. In this work, we introduce the gathering property and show that it is strongly connected with the decryption failure rate of QC-MDPC. Based on the gathering property, we present two results for QC-MDPC based schemes. The first is a new construction of weak keys obtained by extending the keys that have gathering property via ring isomorphism. For the set of weak keys, we present a rigorous analysis of the probability, as well as experimental simulation of the decryption failure rates. Considering BIKE's parameter set targeting $128$-bit security, our result eventually indicates that the average decryption failure rate is lower bounded by $DFR_{avg} \ge 2^{-122.57}$. The second is a key recovery attack against CCA secure QC-MDPC schemes using decryption failures in a multi-target setting. By decrypting ciphertexts with errors satisfying the gathering property, we show that a single decryption failure can be used to identify whether a target's secret key satisfies the gathering property. Then using the gathering property as extra information, we present a modified information set decoding algorithm that efficiently retrieves the target's secret key. For BIKE's parameter set targeting $128$-bit security, a key recovery attack with complexity $2^{119.88}$ can be expected by using extrapolated decryption failure rates.

We show that the Cherbal-Benchetioui key agreement scheme [Comput. Electr. Eng., 109, 108759 (2023)] fails to keep user anonymity, not as claimed. The scheme simply thinks that user anonymity is equivalent to protecting the user's real identity. But the true anonymity means that the adversary cannot attribute different sessions to target entities, which relates to entity-distinguishable, not just identity-revealable.

Learning with Errors (LWE) is a hard math problem used in post-quantum cryptography. Homomorphic Encryption (HE) schemes rely on the hardness of the LWE problem for their security, and two LWE-based cryptosystems were recently standardized by NIST for digital signatures and key exchange (KEM). Thus, it is critical to continue assessing the security of LWE and specific parameter choices. For example, HE uses secrets with small entries, and the HE community has considered standardizing small sparse secrets to improve efficiency and functionality. However, prior work, SALSA and PICANTE, showed that ML attacks can recover sparse binary secrets. Building on these, we propose VERDE, an improved ML attack that can recover sparse binary, ternary, and narrow Gaussian secrets. Using improved preprocessing and secret recovery techniques, VERDE can attack LWE with larger dimensions ($n=512$) and smaller moduli ($\log_2 q=12$ for $n=256$), using less time and power. We propose novel architectures for scaling. Finally, we develop a theory that explains the success of ML LWE attacks.

Classic MLaaS solutions suffer from privacy-related risks since the user is required to send unencrypted data to the server hosting the MLaaS. To alleviate this problem, a thriving line of research has emerged called Privacy-Preserving Machine Learning (PPML) or secure MLaaS solutions that use cryptographic techniques to preserve the privacy of both the input of the client and the output of the server. However, these implementations do not take into consideration the possibility of transferability of known attacks in classic MLaaS settings to PPML settings. In this paper, we demonstrate that it is possible to transfer existing model-extraction attacks using adversarial examples to PPML settings due to relaxed constraints on the abilities of the adversary. We show a working example of an end-to-end attack on an image processing application built using a popular FHE-based framework, namely Concrete-ML. We successfully create a cloned model with just 5000 queries, which is, in fact, 10× less than the size of the training set of the victim model, while incurring only a 7% loss in accuracy. Further, we incorporate the well-known defense strategy against such attacks and show that our attack is still able to clone the model. Finally, we evaluate the different defense rationales that exist in literature and observe that such model stealing attacks are difficult to prevent in secure MLaaS settings.

The partial sums cryptanalytic technique was introduced in 2000 by Ferguson et al., who used it to break 6-round AES with time complexity of $2^{52}$ S-box computations -- a record that has not been beaten ever since. In 2014, Todo and Aoki showed that for 6-round AES, partial sums can be replaced by a technique based on the Fast Fourier Transform (FFT), leading to an attack with a comparable complexity. In this paper we show that the partial sums technique can be combined with an FFT-based technique, to get the best of the two worlds. Using our combined technique, we obtain an attack on 6-round AES with complexity of about $2^{46.4}$ additions. We fully implemented the attack experimentally, along with the partial sums attack and the Todo-Aoki attack, and confirmed that our attack improves the best known attack on 6-round AES by a factor of more than 32. We expect that our technique can be used to significantly enhance numerous attacks that exploit the partial sums technique. To demonstrate this, we use our technique to improve the best known attack on 7-round Kuznyechik by a factor of more than 80, and to reduce the complexity of the best known attack on the full MISTY1 from $2^{69.5}$ to $2^{67}$.

Over the last decades, fault injection attacks have been demonstrated to be an effective method for breaking the security of electronic devices. Some types of fault injection attacks, like clock and voltage glitching, require very few resources by the attacker and are practical and simple to execute. A cost-effective countermeasure against these attacks is the use of a detector circuit which detects timing violations - the underlying effect that glitch attacks rely on. In this paper, we take a closer look at three examples of such detectors that have been presented in the literature. We demonstrate four high-speed clock glitching attacks, which successfully inject faults in systems, where detectors have been implemented to protect. The attacks remain unnoticed by the glitch detectors. We verify our attacks with practical experiments on an FPGA.

Division property is an effective method for finding integral distinguishers for block ciphers, performing cube attacks on stream ciphers, and studying the algebraic degree of boolean functions. One of the main problems in this field is how to provably find the smallest input multiset leading to a balanced output. In this paper, we propose a new method based on division property for finding integral distinguishers with a provably minimum data complexity on permutation functions and block ciphers, in the conventional division property model. The new method is based on efficiently analyzing the algebraic normal form of the target output boolean function. We examine the proposed method on LBlock, TWINE, SIMON, Present, Gift, and Clyde-128 block ciphers. Although in most cases, the results are compliant with the distinguishers reported in the previous work, the proposed method proves the optimality of these results, in the conventional division property model. However, the proposed method can find distinguishers for 8-round Clyde-128 with a data complexity less than the previously reported one, based on conventional division property. The new method is also capable of determining the maximum number of balanced output bits in an integral distinguisher with a specified number of active bits. We propose an algorithm to exploit this capability and apply it to the studied ciphers. As a result, we determine the maximum number of balanced bits on integral distinguishers with minimum and non-minimum data complexities on the studied ciphers and report improved results on Gift-64, Present and SIMON64 in the conventional model.

This paper re-analyzes the algorithm proposed by Guedes, Assis, and Lula in 2012, which they claimed that the algorithm can break Blum-Micali Pseudorandom number generator in polynomial time. We used a 5×5 transformation matrix instead of the original 2×2 transformation matrix, which can include terms that they missed in their analysis. We proved that their proposed algorithm cannot break the pseudorandom number generator, because during the amplitude amplification process, two iterations of the circuit is the same as an identity gate. To solve this problem, we proposed a corrected algorithm based on Grover’s Search Algorithm for NP-complete problems, which breaks the Blum-Micali Pseudorandom number generator in \(\mathcal{O}(n^4 2^{n/2})\). We conclude that the Blum-Micali Pseudorandom number generator is still quantum resistant. This study indicates that the discrete logarithm problem and prime factorization could still be the foundations of quantum-resistant cryptographical applications.

We show that the Xu et al.'s authentication and key agreement scheme [IEEE Trans. Ind. Informatics, 18(10), 7118-7127, 2022] is flawed. (1) It confused some operations for bilinear maps and presented some inconsistent computations. (2) It failed to keep anonymity, not as claimed. The adversary can use any device's public key stored in the blockchain to test some verification equations so as to reveal the identity of a target device.

Power analysis of public-key algorithms is a well-known approach in the community of side-channel analysis. We usually classify operations based on the differences in power traces produced by different basic operations (such as modular exponentiation) to recover secret information like private keys. The more accurate the segmentation of power traces, the higher the efficiency of their classification. There exist two commonly used methods: one is equidistant segmentation, which requires a fixed number of basic operations and similar trace lengths for each type of operation, leading to limited application scenarios; the other is peak-based segmentation, which relies on personal experience to configure parameters, resulting in insufficient flexibility and poor universality. In this paper, we propose an automated power trace segmentation method based on reinforcement learning algorithms, which is applicable to a wide range of common implementation of public-key algorithms. Reinforcement learning is an unsupervised machine learning technique that eliminates the need for manual label collection. For the first time, this technique is introduced into the field of side-channel analysis for power trace processing. By using prioritized experience replay optimized Deep Q-Network algorithm, we reduce the number of parameters required to achieve accurate segmentation of power traces to only one, i.e. the key length. We also employ various techniques to improve the segmentation effectiveness, such as clustering algorithm, enveloped-based feature enhancement and fine-tuning method. We validate the effectiveness of the new method in nine scenarios involving hardware and software implementations of different public-key algorithms executed on diverse platforms such as microcontrollers, SAKURA-G, and smart cards. Specifically, one of these implementations is protected by time randomization countermeasures. Experimental results show that our method has good robustness on the traces with varying segment lengths and differing peak heights. After employ the clustering algorithm, our method achieves an accuracy of over 99.6% in operations recovery. Besides, power traces collected from these devices have been uploaded as databases, which are available for researchers engaged in public-key algorithms to conduct related experiments or verify our method.

In an effort to circumvent the high cost of standard countermeasures against side-channel attacks in post-quantum cryptography, some works have developed low-cost detection-based countermeasures. These countermeasures try to detect maliciously generated input ciphertexts and react to them by discarding the ciphertext or secret key. In this work, we take a look at two previously proposed low-cost countermeasures: the ciphertext sanity check and the decapsulation failure check, and demonstrate successful attacks on these schemes. We show that the first countermeasure can be broken with little to no overhead, while the second countermeasure requires a more elaborate attack strategy that relies on valid chosen ciphertexts. Thus, in this work, we propose the first chosen-ciphertext based side-channel attack that only relies on valid ciphertexts for key recovery. As part of this attack, a third contribution of our paper is an improved solver that retrieves the secret key from linear inequalities constructed using side-channel leakage from the decryption procedure. Our solver is an improvement over the state-of-the-art Belief Propagation solvers by Pessl and Prokop, and later Delvaux. Our method is simpler, easier to understand and has lower computational complexity, while needing less than half the inequalities compared to previous methods.

Lattice reduction algorithms such as BKZ (Block-Korkine-Zolotarev) play a central role in estimating the security of lattice-based cryptography. The subroutine in BKZ which finds the shortest vector in a projected sublattice can be instantiated with enumeration algorithms. The enumeration procedure can be seen as a depth-first search on some ``enumeration tree'' whose nodes denote a partial assignment of the coefficients, corresponding to lattice points as a linear combination of the lattice basis with the coefficients. This work provides a concrete analysis for the cost of quantum lattice enumeration based on Montanaro's quantum tree backtracking algorithm. More precisely, we give a concrete implementation in the quantum circuit model. We also show how to optimize the circuit depth by parallelizing the components. Based on the circuit designed, we discuss the concrete quantum resource estimates required for lattice enumeration.

In this work, we propose a novel single-trace key recovery attack targeting side-channel leakage from the key-generation and encryption procedure of Kyber KEM. Our attack exploits the inherent nature of the Module-Learning With Errors (Module-LWE) problem used in Kyber KEM. We demonstrate that the inherent reliance of Kyber KEM on the Module-LWE problem results in higher number of repeated and secret key-related computations, referred to as STAMPs appearing on a single side channel trace, compared to the Ring-LWE problem of similar security level. We exploit leakage from the pointwise multiplication operation and take advantage of the properties of the Module-LWE instance to enable a potential single trace key recovery attack. We validated the efficacy of our attack on both simulated and real traces, and we performed experiments on both the reference and assembly optimized implementation of Kyber KEM, taken from the pqm4 library, a well-known benchmarking and testing framework for PQC schemes on the ARM Cortex-M4 microcontroller. We also analyze the applicability of our attack on countermeasures against traditional SCA such as masking and shuffling. We believe our work motivates more research towards SCA resistant implementation of key-generation and encryption procedure for Kyber KEM.

This paper formally analyzes two major non-profiled deep-learning-based side-channel attacks (DL-SCAs): differential deep-learning analysis (DDLA) by Timon and collision DL-SCA by Staib and Moradi. These DL-SCAs leverage supervised learning in non-profiled scenarios. Although some intuitive descriptions of these DL-SCAs exist, their formal analyses have been rarely conducted yet, which makes it unclear why and when the attacks succeed and how the attack can be improved. In this paper, we provide the first information-theoretical analysis of DDLA. We reveal its relevance to the mutual information analysis (MIA), and then present three theorems stating some limitations and impossibility results of DDLA. Subsequently, we provide the first probability-theoretical analysis on collision DL-SCA. After presenting its formalization with a proposal of our distinguisher for collision DL-SCA, we prove its optimality. Namely, we prove that the collision DL-SCA using our distinguisher theoretically maximizes the success rate if the neural network (NN) training is completely successful (namely, the NN completely imitates the true conditional probability distribution). Accordingly, we propose an improvement of the collision DL-SCA based on a dedicated NN architecture and a full-key recovery methodology using multiple neural distinguishers. Finally, we experimentally evaluate non-profiled (DL-)SCAs using a newly created dataset using publicly available first-order masked AES implementation. The existing public dataset of side-channel traces is insufficient to evaluate collision DL-SCAs due to a lack of substantive side-channel traces for different key values. Our dataset enables a comprehensive evaluation of collision (DL-)SCAs, which clarifies the current situation of non-profiled (DL-)SCAs.

Masking, as a common countermeasure, has been widely utilized to protect cryptographic implementations against power side-channel attacks. It significantly enhances the difficulty of attacks, as the sensitive intermediate values are randomly partitioned into multiple parts and executed on different times. The adversary must amalgamate information across diverse time samples before launching an attack, which is generally accomplished by feature extraction (e.g., Points-Of-Interest (POIs) combination and dimensionality reduction). However, traditional POIs combination methods, machine learning and deep learning techniques are often too time consuming, and necessitate a significant amount of computational resources. In this paper, we undertake the first study on manifold learning and their applications against masked cryptographic implementations. The leaked information, which manifests as the manifold of high-dimensional power traces, is mapped into a low-dimensional space and achieves feature extraction through manifold learning techniques like ISOMAP, Locally Linear Embedding (LLE), and Laplacian Eigenmaps (LE). Moreover, to reduce the complexity, we further construct explicit polynomial mappings for manifold learning to facilitate the dimensionality reduction. Compared to the classical machine learning and deep learning techniques, our schemes built from manifold learning techniques are faster, unsupervised, and only require very simple parameter tuning. Their effectiveness has been fully validated by our detailed experiments.

Ascon is a recently standardized suite of symmetric cryptography for authenticated encryption and hashing algorithms designed to be lightweight. The Ascon scheme has been studied since it was introduced in 2015 for the CAESAR competition, and many efforts have been made to transform this hardware-oriented scheme to work with any embedded device architecture. Ascon is designed with side-channel resistance in mind and can also be protected with countermeasures against side-channel analysis. Up to now, the effort of side-channel analysis is mainly put on hardware implementations, with only a few studies being published on the real-world side-channel security of software implementations. In this paper, we give a comprehensive view of the side-channel security of Ascon implemented on a 32-bit microcontroller for both the reference and a protected implementation. We show different potential leakage functions that can lead to real-world leakages and demonstrate the most potent attacks that can be obtained with the corresponding leakage functions. We present our results using correlation power analysis (CPA) and deep learning-based side-channel analysis and provide a practical estimation of the efforts needed for an attacker to recover the complete key used for authenticated encryption. Our results show that the reference implementation is not side-channel secure since an attacker can recover the full key with 8,000 traces using CPA and around 1,000 traces with deep learning analysis. While second-order CPA cannot recover any part of the secret, deep learning side-channel analysis can recover partial keys with 800 traces on the protected implementation. Unfortunately, the model used for multi-task key recovery lacks the generalization to correctly recover all partial keys for the full key attack.

The emergence of hardware trojans as significant threats in various aspects of hardware design, including Firmware, open-source IP, and PCB design, has raised serious concerns. Simultaneously, AI technologies have been employed to simplify the complexity of Side Channel Analysis (SCA) attacks. Due to the increasing risk posed by these threats, it becomes essential to test hardware by considering all possible attack vectors. This paper aims to propose a black box attack using side channel analysis with the aid of hardware trojan insertion. The objective is to emphasize the necessity of side channel-based testing to defend against such attacks. The proposed attack can be executed in FPGA, ASIC, and microcontroller designs. The paper is primarily focused on Verilog design based hardware trojan insertion, and a small example demonstration is provided to illustrate this attack. Keywords:

We show that the key agreement scheme [Quantum Inf. Process., 20:188, 2021] is flawed. (1) It requires that the quantum channel must be intact so as to keep the transferred photon sequences complete and undamaged, even if the channel is tapped. But this is unrealistic because of quantum non-cloning theorem. (2) The user's capability is artificially assumed, who can measure a hybrid photon sequence only with $Z$-basis, unable to measure with $X$-basis. (3) It requires an authenticated classical channel for the negotiation between Alice and Server$_B$. If such a channel is available, the scheme can be greatly simplified using the mechanism in BB84 protocol.

Blind signatures allow the issuing of signatures on messages chosen by the user so that they ensure $\mathit{blindness}$ of the message against the signer. Moreover, a malicious user cannot output $\ell+1$ signatures while only finishing $\ell$ signing sessions. This notion, called $\mathit{one}$-$\mathit{more}$ unforgeability, comes in two flavors supporting either $\mathit{sequential}$ or $\mathit{concurrent}$ sessions. In this paper, we investigate the security of a class of blind signatures constructed from Sigma-protocols with small challenge space $\mathcal{C}_{\Sigma}$ (i.e., polynomial in the security parameter), using $k$ repetitions of the protocol to decrease the chances of a cheating prover. This class of schemes includes, among others, the Schnorr blind signature scheme with bit challenges and the recently proposed isogeny-based scheme CSI-Otter (Crypto'23), as well as potential blind signatures designed from assumptions with the well-known Sigma-protocol for the graph-isomorphism problem (e.g., Lattice Isomorphism Problem). For this class of blind signatures, we show a $\mathit{polynomial}$-$\mathit{time}$ attack that breaks one-more unforgeability for any $\ell \geq k$ concurrent sessions in time $O(k \cdot |\mathcal{C}_{\Sigma}|)$. Contrary to the ROS attack, ours is generic and does not require any particular algebraic structure. We also propose a computational trade-off, where, for any $t \leq k$, our attack works for $\ell = \frac{k}{t}$ in time $O(\frac{k}{t} \cdot |\mathcal{C}_{\Sigma}|^t)$. The consequences of our attack are as follows. Schemes in the investigated class of blind signatures should not be used concurrently without applying specific transformations to boost the security to support more signing sessions. Moreover, for the parameters proposed for CSI-Otter ($k=128$ and $|\mathcal{C}_{\Sigma}|=2$), the scheme becomes forgeable after 128 concurrent signing sessions for the basic attack and with only eight sessions in our optimized attack. We also show that for those parameters, it is even possible to compute two signatures in around 10 minutes with just one signing session using the computation power of the Bitcoin network. Thus, we show that, for sequential security, the parameter $k$ must be at least doubled in the security parameter for any of the investigated schemes.

Hash-based signature (HBS) schemes are an efficient method of guaranteeing the authenticity of data in a post-quantum world. The stateful schemes LMS and XMSS and the stateless scheme SPHINCS+ are already standardised or will be in the near future. The Winternitz one-time signature (WOTS) scheme is one of the fundamental building blocks used in all these HBS standardisation proposals. We present a new fault injection attack targeting WOTS that allows an adversary to forge signatures for arbitrary messages. The attack affects both the signing and verification processes of all current stateful and stateless schemes. Our attack renders the checksum calculation within WOTS useless. A successful fault injection allows at least an existential forgery attack and, in more advanced settings, a universal forgery attack. While checksum computation is clearly a critical point in WOTS, and thus in any of the relevant HBS schemes, its resilience against a fault attack has never been considered. To fill this gap, we theoretically explain the attack, estimate its practicability, and derive the brute-force complexity to achieve signature forgery for a variety of parameter sets. We analyse the reference implementations of LMS, XMSS and SPHINCS+ and pinpoint the vulnerable points. To harden these implementations, we propose countermeasures and evaluate their effectiveness and efficiency. Our work shows that exposed devices running signature generation or verification with any of these three schemes must have countermeasures in place.

In cube attacks, key filtering is a basic step of identifying the correct key candidates by referring to the truth tables of superpolies. When terms of superpolies get massive, the truth table lookup complexity of key filtering increases significantly. In this paper, we propose the concept of implementation dependency dividing all cube attacks into two categories: implementation dependent and implementation independent. The implementation dependent cube attacks can only be feasible when the assumption that one encryption oracle query is more complicated than one table lookup holds. On the contrary, implementation independent cube attacks remain feasible in the extreme case where encryption oracles are implemented in the full codebook manner making one encryption query equivalent to one table lookup. From this point of view, we scrutinize existing cube attack results of stream ciphers Trivium, Grain-128AEAD, Acorn and Kreyvium. As a result, many of them turn out to be implementation dependent. Combining with the degree evaluation and divide-and-conquer techniques used for superpoly recovery, we further propose new cube attack results on Kreyvium reduced to 898, 899 and 900 rounds. Such new results not only mount to the maximal number of rounds so far but also are implementation independent.

Quantum computers, especially those with over 10,000 qubits, pose a potential threat to current public key cryptography systems like RSA and ECC due to Shor's algorithms. Grover's search algorithm is another quantum algorithm that could significantly impact current cryptography, offering a quantum advantage in searching unsorted data. Therefore, with the advancement of quantum computers, it is crucial to analyze potential quantum threats. While many works focus on Grover’s attacks in symmetric key cryptography, there has been no research on the practical implementation of the quantum approach for lattice-based cryptography. Currently, only theoretical analyses involve the application of Grover's search to various Sieve algorithms. In this work, for the first time, we present a quantum NV Sieve implementation to solve SVP, posing a threat to lattice-based cryptography. Additionally, we implement the extended version of the quantum NV Sieve (i.e., the dimension and rank of the lattice vector). Our extended implementation could be instrumental in extending the upper limit of SVP (currently, determining the upper limit of SVP is a vital factor). Lastly, we estimate the quantum resources required for each specific implementation and the application of Grover's search. In conclusion, our research lays the groundwork for the quantum NV Sieve to challenge lattice-based cryptography. In the future, we aim to conduct various experiments concerning the extended implementation and Grover's search.

At ASIACRYPT 2019, Zhang proposed a near collision attack on A5/1 claiming to recover the 64-bit A5/1 state with a time complexity around $2^{32}$ cipher ticks with negligible memory requirements. Soon after its proposal, Zhang's near collision attack was severely challenged by Derbez \etal who claimed that Zhang's attack cannot have a time complexity lower than Golic's memoryless guess-and-determine attack dating back to EUROCRYPT 1997. In this paper, we study both the guess-and-determine and the near collision attacks for recovering A5/1 states with negligible memory complexities. Firstly, we propose a new guessing technique called the \emph{move guessing technique} that can construct linear equation filters in a more efficient manner. Such a technique can be applied to both guess-and-determine and collision attacks for efficiency improvements. Secondly, we take the filtering strength of the linear equation systems into account for complexity analysis. Such filtering strength are evaluated with practical experiments making the complexities more convincing. Based on such new techniques, we are able to give 2 new guess-and-determine attacks on A5/1: the 1st attack recovers the internal state $\vec{s}^0$ with time complexity $2^{43.92}$; the 2nd one recovers a different state $\vec{s}^1$ with complexity $2^{43.25}$. We also revisit Golic's guess-and-determine attack and Zhang's near collision attacks. According to our detailed analysis, the complexity of Golic's $\vec{s}^1$ recovery attack is no lower than $2^{46.04}$, higher than the previously believed $2^{43}$. On the other hand, Zhang's near collision attack recovers $\vec{s}^0$ with the time complexity $2^{53.19}$: such a complexity can be further lowered to $2^{50.78}$ with our move guessing technique.

The problem of decoding random codes is a fundamental problem for code-based cryptography, including recent code-based candidates in the NIST post-quantum standardization process. In this paper, we present a novel sieving-style information-set decoding (ISD) algorithm, addressing the task of solving the syndrome decoding problem. Our approach involves maintaining a list of weight-$2p$ solution vectors to a partial syndrome decoding problem and then creating new vectors by identifying pairs of vectors that collide in $p$ positions. By gradually increasing the parity-check condition by one and repeating this process iteratively, we find the final solution(s). We show that our novel algorithm performs better than other ISDs in the memory-restricted scenario when applied to McEliece. Notably, in the case of problem instances with very low relative weight, it seems to significantly outperform all previous algorithms. In particular, for code-based candidate BIKE, the algorithm has lower bit complexity than the previous best results.

We show that the scheme [Neurocomputing, 2022 (500), 741-749] fails to keep anonymity, not as claimed. The scheme neglects the basic requirement for bit-wise XOR, and tries to encrypt data by the operator. The negligence results in some trivial equalities. An adversary can retrieve the user's identity from one captured string via the open channel.

A new batch of ``complete and proper'' digital signature scheme submissions has recently been published by NIST as part of its process for establishing post-quantum cryptographic standards. This note communicates an attack on the 3WISE digital signature scheme that the submitters did not wish to withdraw after NIST communicated it to them. While the 3WISE digital signature scheme is based on a collection of cubic maps which are naturally modeled as symmetric 3-tensors and 3-tensor rank is a difficult problem, the multivariate signature scheme is still vulnerable to MinRank attacks upon projection. We are able to break the NIST security level I parameters within a few seconds. Since the attack is polynomial time, there is no reparametrization resulting in a secure scheme.

Recently, there has been a surge of interest in the security of authenticated encryption with associated data (AEAD) within the context of key commitment frameworks. Security within this framework ensures that a ciphertext chosen by an adversary does not decrypt to two different sets of key, nonce, and associated data. Despite this increasing interest, the security of several widely deployed AEAD schemes has not been thoroughly examined within this framework. In this work, we assess the key committing security of AEGIS, which emerged as a winner in the Competition for Authenticated Encryption: Security, Applicability, and Robustness (CAESAR). A recent assertion has been made suggesting that there are no known attacks on AEGIS in the key committing settings and AEGIS qualifies as a fully committing AEAD scheme in IETF document. However, contrary to this claim, we propose a novel O(1) attack applicable to all variants of AEGIS. This demonstrates the ability to execute a key committing attack within the FROB game setting, which is known to be one of the most stringent key committing frameworks. This implies that our attacks also hold validity in other, more relaxed frameworks, such as CMT-1, CMT-4, and so forth.

In this paper, we provide experimental evidence for the benefits of multi-task learning in the context of masked AES implementations (via the ASCADv1-r and ASCADv2 databases). We develop an approach for comparing single-task and multi-task approaches rather than comparing specific resulting models: we do this by training many models with random hyperparameters (instead of comparing a few highly tuned models). We find that multi-task learning has significant practical advantages that make it an attractive option in the context of device evaluations: the multi-task approach leads to performant networks quickly in particular in situations where knowledge of internal randomness is not available during training.

We show that the key agreement scheme [J. Supercomput., 78:12093-12113, 2022] is flawed. (1) It neglects the representation of a point over an elliptic curve and the basic requirement for bit-wise XOR, which results in a trivial equality. By the equality, an adversary can recover a target device's identity, which means the scheme fails to keep anonymity. (2) It falsely requires that the central server should share its master secret key with each dew server. (3) The specified certificate is almost nonsensical.

Exclusive OR (XOR), a common Boolean logical operation, is an operation on two factors where the result is true if and only if one operand is true and the other is false. A simple way to state this is ``one or the other, but not both''. Using this logical operation, a text string can be encrypted by applying the XOR operator to every character using a ``key''. If you want to decrypt the output, simply reapply the key and the resulting output will be the original message.

Despite considerable achievements of deep learning-based side-channel analysis, overfitting represents a significant obstacle in finding optimized neural network models. This issue is not unique to the side-channel domain. Regularization techniques are popular solutions to overfitting and have long been used in various domains. At the same time, the works in the side-channel domain show sporadic utilization of regularization techniques. What is more, no systematic study investigates these techniques' effectiveness. In this paper, we aim to investigate the regularization effectiveness on a randomly selected model, by applying four powerful and easy-to-use regularization techniques to eight combinations of datasets, leakage models, and deep learning topologies. The investigated techniques are $L_1$, $L_2$, dropout, and early stopping. Our results show that while all these techniques can improve performance in many cases, $L_1$ and $L_2$ are the most effective. Finally, if training time matters, early stopping is the best technique.

We conduct a security analysis of the e-voting protocol used for the largest political election using e-voting in the world, the 2022 French legislative election for the citizens overseas. Due to a lack of system and threat model specifications, we built and contributed such specifications by studying the French legal framework and by reverse-engineering the code base accessible to the voters. Our analysis reveals that this protocol is affected by two design-level and implementation-level vulnerabilities. We show how those allow a standard voting server attacker and even more so a channel attacker to defeat the election integrity and ballot privacy due to 6 attack variants. We propose and discuss 5 fixes to prevent those attacks. Our specifications, the attacks, and the fixes were acknowledged by the relevant stakeholders during our responsible disclosure. Our attacks are in the process of being prevented with our fixes for future elections. Beyond this specific protocol, we draw general conclusions and lessons from this instructive experience where an e-voting protocol meets the real-world constraints of a large-scale and political election.

The well-defined information leakage of Encrypted Search Algorithms (ESAs) is predominantly analyzed by crafting so-called leakage attacks. These attacks utilize adversarially known auxiliary data and the observed leakage to attack an ESA instance built on a user's data. Known-data attacks require the auxiliary data to be a subset of the user's data. In contrast, sampled-data attacks merely rely on auxiliary data that is, in some sense, statistically close to the user's data and hence reflect a much more realistic attack scenario where the auxiliary data stems from a publicly available data source instead of the private user's data. Unfortunately, it is unclear what "statistically close" means in the context of sampled-data attacks. This leaves open how to measure whether data is close enough for attacks to become a considerable threat. Furthermore, sampled-data attacks have so far not been evaluated in the more realistic attack scenario where the auxiliary data stems from a source different to the one emulating the user's data. Instead, auxiliary and user data have been emulated with data from one source being split into distinct training and testing sets. This leaves open whether and how well attacks work in the mentioned attack scenario with data from different sources. In this work, we address these open questions by providing a measurable metric for statistical closeness in encrypted keyword search. Using real-world data, we show a clear exponential relation between our metric and attack performance. We uncover new data that are intuitively similar yet stem from different sources. We discover that said data are not "close enough" for sampled-data attacks to perform well. Furthermore, we provide a re-evaluation of sampled-data keyword attacks with varying evaluation parameters and uncover that some evaluation choices can significantly affect evaluation results.

GIFT is a family of lightweight block ciphers based on SPN structure and composed of two versions named GIFT-64 and GIFT-128. In this paper, we reevaluate the security of GIFT-64 against the rectangle attack under the related-key setting. Investigating the previous rectangle key recovery attack on GIFT-64, we obtain the core idea of improving the attack——trading off the time complexity of each attack phase. We flexibly guess part of the involved subkey bits to balance the time cost of each phase so that the overall time complexity of the attack is reduced. Moreover, the reused subkey bits are identified according to the linear key schedule of GIFT-64 and bring additional advantages for our attacks. Furthermore, we incorporate the above ideas and propose a dedicated MILP model for finding the best rectangle key recovery attack on GIFT-64. As a result, we get the improved rectangle attacks on 26-round GIFT-64, which are the best attacks on it in terms of time complexity so far.

Dual attacks aiming at decoding generic linear codes have been found recently to outperform for certain parameters information set decoding techniques which have been for $60$ years the dominant tool for solving this problem and choosing the parameters of code-based cryptosystems. However, the analysis of the complexity of these dual attacks relies on some unproven assumptions that are not even fully backed up with experimental evidence. These dual attacks can actually be viewed as the code-based analogue of dual attacks in lattice based cryptography. Here too, dual attacks have been found out those past years to be strong competitors to primal attacks and a controversy has emerged whether similar heuristics made for instance on the independence of certain random variables really hold. We will show that the dual attacks in coding theory can be studied by providing in a first step a simple alternative expression of the fundamental quantity used in these dual attacks. We then show that this expression can be studied without relying on independence assumptions whatsoever. This study leads us to discover that there is indeed a problem with the latest and most powerful dual attack introduced in the recent Asiacrypt 2022 paper "Statistical decoding 2.0: Reducing Decoding to LPN" (RLPN) and that for the parameters chosen in this algorithm there are indeed false candidates which are produced and which are not predicted by the analysis provided there which relies on independence assumptions. We then suggest a slight modification of this algorithm consisting in a further verification step, analyze it thoroughly, provide experimental evidence that our analysis is accurate and show that the complexity claims made in RLPN are indeed valid for this modified algorithm. This approach provides a simple methodology for studying rigorously dual attacks which could turn out to be useful for further developing the subject.

The support minors method has become indispensable to cryptanalysts in attacking various post-quantum cryptosystems in the areas of multivariate cryptography and rank-based cryptography. The complexity analysis for support minors minrank calculations is a bit messy, with no closed form for the Hilbert series of the ideal generated by the support minors equations (or, more correctly, for the quotient of the polynomial ring by this ideal). In this article, we provide a generating series whose coefficients are the Hilbert Series of related MinRank ideals. This simple series therefore reflects and relates the structure of all support minors ideals. Its simplicity also makes it practically useful in computing the complexity of support minors instances.

Ascon, a family of algorithms that supports hashing and authenticated encryption, is the winner of the NIST Lightweight Cryptography Project. In this paper, we propose an improved preimage attack against 2-round Ascon-XOF-64 with a complexity of $2^{32}$ via a better guessing strategy. Furthermore, in order to find a good guessing strategy efficiently, we build a MILP model and successfully extend the attack to 3 rounds. The time complexity is $2^{53}$ when $IV=0$, while for the real $IV$, the attack still works and the time complexity is $2^{51}$. Additionally, we also investigate the resistance of Ascon-HASH against collision attacks. We introduce the linearization of the inverse of S-boxes and then propose a practical free-start collision attack on 3-round Ascon-HASH using a differential trail searched dedicatedly. Furthermore, We construct different 2-round connectors using the linearization of the inverse of S-boxes and successfully extend the collision attack to 4 rounds and 5 rounds of Ascon-HASH with complexities of $2^{21}$ and $2^{41}$ respectively. Although our attacks do not compromise the security of the full 12-round Ascon-XOF and Ascon-HASH, they provide some insights into Ascon's security.

Gentry's groundbreaking work showed that a fully homomorphic, provably secure scheme is possible via bootstrapping a somewhat homomorphic scheme. However, a major drawback of bootstrapping is its high computational cost. One alternative is to use a different metric for noise so that homomorphic operations do not accumulate noise, eliminating the need for boostrapping altogether. Leonardi and Ruiz-Lopez present a group-theoretic framework for such a ``noise non-accumulating'' multiplicative homomorphic scheme, but Agathocleous et al. expose weaknesses in this framework when working over finite abelian groups. Tangentially, Li and Wang present a ``noise non-accumulating'' fully homomorphic scheme by performing Ostrovsky and Skeith's transform on a multiplicative homomorphic scheme of non-abelian group rings. Unfortunately, the security of Li and Wang's scheme relies on the Factoring Large Numbers assumption, which is false given an adversary with a quantum computer. In this work, we seek to modify Li and Wang's scheme to be post-quantum secure by fitting it into the Leonardi and Ruiz-Lopez framework for non-abelian rings. We discuss improved security assumptions for Li and Wang encryption and assess the shortcomings of working in a non-abelian setting. Finally, we show that a large class of semisimple rings is incompatible with the Leonardi and Ruiz-Lopez framework.

The meet-in-the-middle (MITM) technique has led to many key-recovery attacks on block ciphers and preimage attacks on hash functions. Nowadays, cryptographers use automatic tools that reduce the search of MITM attacks to an optimization problem. Bao et al. (EUROCRYPT 2021) introduced a low-level modeling based on Mixed Integer Linear Programming (MILP) for MITM attacks on hash functions, which was extended to key-recovery attacks by Dong et al. (CRYPTO 2021). However, the modeling only covers AES-like designs. Schrottenloher and Stevens (CRYPTO 2022) proposed a different approach aiming at higher-level simplified models. However, this modeling was limited to cryptographic permutations. In this paper, we extend the latter simplified modeling to also cover block ciphers with simple key schedules. The resulting modeling enables us to target a large array of primitives, typically lightweight SPN ciphers where the key schedule has a slow diffusion, or none at all. We give several applications such as full breaks of the PIPO-256 and FUTURE block ciphers, and reduced-round classical and quantum attacks on SATURNIN-Hash.

In this paper we show that Bleichenbacher-style attacks on RSA decryption are not only still possible, but also that vulnerable implementations are common. We have successfully attacked multiple implementations using only timing of decryption operation and shown that many others are vulnerable. To perform the attack we used more statistically rigorous techniques like the sign test, Wilcoxon signed-rank test, and bootstrapping of median of pairwise differences. We publish a set of tools for testing libraries that perform RSA decryption against timing side-channel attacks, including one that can test arbitrary TLS servers with no need to write a test harnesses. Finally, we propose a set of workarounds that implementations can employ if they can't avoid the use of RSA.

In this paper we analyse typical timing data that can be collected over loopback interface, in local, and in metropolitan area networks. We evaluate performance of few statistical test for detecting differences in timing of server responses. The evaluated tests include the popular Box test, as well as sign test, Wilcoxon signed-rank test, and paired sample t-test. We found that the Box test offers poor performance, as it's an incorrect test to use for the measurements we collected. Use of appropriate tests also allows for robust differentiation between much smaller differences than the existing literature would suggest. We were able to detect side channels of single-digit CPU cycles over regular gigabit Ethernet. Those alternative tests were also found to be robust against noise in production networks, allowing detection of side channel of just few nanoseconds with 6 network hops between test systems.

Thangavel and Varalakshmi proposed an enhanced DNA and ElGamal cryptosystem for secure data storage and retrieval in cloud. They modified ElGamal algorithm which it calls enhanced ElGamal cryptosystem. We prove that their enhanced ElGamal scheme, which does not require two random numbers by data owner. Although the attacker is unable to find out what message the data owner gave to the data user. However, the attackers can still confuse the issue of sending messages to data users. On the other hand, this scheme can not against insider attack, therefore it is insecure.

Elisabeth-4 is a stream cipher tailored for usage in hybrid homomorphic encryption applications that has been introduced by Cosseron et al. at ASIACRYPT 2022. In this paper, we present several variants of a key-recovery attack on the full Elisabeth-4 that break the 128-bit security claim of that cipher. Our most optimized attack is a chosen-IV attack with a time complexity of $2^{88}$ elementary operations, a memory complexity of $2^{54}$ bits and a data complexity of $2^{41}$ bits. Our attack applies the linearization technique to a nonlinear system of equations relating some keystream bits to the key bits and exploits specificities of the cipher to solve the resulting linear system efficiently. First, due to the structure of the cipher, the system to solve happens to be very sparse, which enables to rely on sparse linear algebra and most notably on the Block Wiedemann algorithm. Secondly, the algebraic properties of the two nonlinear ingredients of the filtering function cause rank defects which can be leveraged to solve the linearized system more efficiently with a decreased data and time complexity. We have implemented our attack on a toy version of Elisabeth-4 to verify its correctness. It uses the efficient implementation of the Block Wiedemann algorithm of CADO-NFS for the sparse linear algebra.

We show that the key agreement scheme [J. Syst. Archit., 131:102698, 2022] fails to keep user anonymity and service provider anonymity, not as claimed. The scheme simply thinks that user anonymity is equivalent to protecting the target user's identity against exposure, while its long-term pseudo-identity can be exposed. We want to clarify that the true anonymity means that an adversary cannot attribute different sessions to different target users, even if the true identifier cannot be retrieved from the exposed pseudo-identifier.

A recent series of works (Hecht, IACR ePrint, 2020–2021) propose to build post-quantum public-key encapsulation, digital signatures, group key agreement and oblivious transfer from "R-propped" variants of the Symmetrical Decomposition and Discrete Logarithm problems for matrix groups over $\mathbb{F}_{2^8}$. We break all four proposals by presenting a linearisation attack on the Symmetrical Decomposition platform, a forgery attack on the signature scheme, and a demonstration of the insecurity of the instances of the Discrete Logarithm Problem used for signatures, group key agreement and oblivious transfer, showing that none of the schemes provides adequate security.

All modern lattice-based schemes build on variants of the LWE problem. Information leakage of the LWE secret $\mathbf{s} \in \mathbb{Z}_q^n$ is usually modeled via so-called hints, i.e., inner products of $\mathbf{s}$ with some known vector. At Crypto`20, Dachman-Soled, Ducas, Gong and Rossi (DDGR) defined among other so-called perfect hints and modular hints. The trailblazing DDGR framework allows to integrate and combine hints successively into lattices, and estimates the resulting LWE security loss. We introduce a new methodology to integrate and combine an arbitrary number of perfect and modular in a single stroke. As opposed to DDGR's, our methodology is significantly more efficient in constructing lattice bases, and thus easily allows for a large number of hints up to cryptographic dimensions -- a regime that is currently impractical in DDGR's implementation. The efficiency of our method defines a large LWE parameter regime, in which we can fully carry out attacks faster than DDGR can solely estimate them. The benefits of our approach allow us to practically determine which number of hints is sufficient to efficiently break LWE-based lattice schemes in practice. E.g., for mod-$q$ hints, i.e., modular hints defined over $\mathbb{Z}_q$, we reconstruct \Kyber-512 secret keys via LLL reduction (only!) with an amount of $449$ hints. Our results for perfect hints significantly improve over these numbers, requiring for LWE dimension $n$ roughly $n/2$ perfect hints. E.g., we reconstruct via LLL reduction \Kyber-512 keys with merely $234$ perfect hints. If we resort to stronger lattice reduction techniques like BKZ, we need even fewer hints. For mod-$q$ hints our method is extremely efficient, e.g., taking total time for constructing our lattice bases and secret key recovery via LLL of around 20 mins for dimension 512. For perfect hints in dimension 512, we require around 3 hours. Our results demonstrate that especially perfect hints are powerful in practice, and stress the necessity to properly protect lattice schemes against leakage.

The differential-linear attack is one of the most effective attacks against ARX ciphers. However, two technical problems are preventing it from being more effective and having more applications: (1) there is no efficient method to search for good differential-linear approximations. Existing methods either have many constraints or are currently inefficient. (2) partitioning technique has great potential to reduce the time complexity of the key-recovery attack, but there is no general tool to construct partitions for ARX ciphers. In this work, we step forward in solving the two problems. First, we propose a novel idea for generating new good differential-linear approximations from known ones, based on which new searching algorithms are designed. Second, we propose a general tool named partition tree, for constructing partitions for ARX ciphers. Based on these new techniques, we present better attacks for two ISO/IEC standards, i.e., LEA and Speck. For LEA, we present the first 17-round distinguisher which is 1 round longer than the previous best distinguisher. Furthermore, we present the first key recovery attacks on 17-round LEA-128, 18-round LEA-192, and 18-round LEA-256, which attack 3, 4, and 3 rounds more than the previous best attacks. For Speck, we find better differential-linear distinguishers for Speck48 and Speck64. The first differential-linear distinguishers for Speck96 and Speck128 are also presented.

In this paper, we improve the cube attack by exploiting low-degree factors of the superpoly w.r.t. certain "special" index set of cube (ISoC). This can be viewed as a special case of the correlation cube attack proposed at Eurocrypt 2018, but under our framework more beneficial equations on the key variables can be obtained in the key-recovery phase. To mount our attack, one has two challenging problems: (1) effectively recover algebraic normal form of the superpoly and extract out its low-degree factors; and (2) efficiently search a large quantity of good ISoCs. We bring in new techniques to solve both of them. First, we propose the variable substitution technique for middle rounds of a cipher, in which polynomials on the key variables in the algebraic expressions of internal states are substituted by new variables. This will improve computational complexity of the superpoly recovery and promise more compact superpolys that can be easily decomposed with respect to the new variables. Second, we propose the vector numeric mapping technique, which seeks out a tradeoff between efficiency of the numeric mapping technique (Crypto 2019) and accuracy of the monomial prediction technique (Asiacrypt 2020) in degree evaluation of superpolys. Combining with this technique, a fast pruning method is given and modeled by MILP to filter good ISoCs of which the algebraic degree satisfies some fixed threshold. Thanks to automated MILP solvers, it becomes practical to comprehensively search for good cubes across the entire search space. To illustrate the power of our techniques, we apply all of them to Trivium stream cipher. As a result, we have recovered the superpolys of three cubes given by Kesarwani et al. in 2020, only to find they do not have zero-sum property up to 842 rounds as claimed in their paper. To our knowledge, the previous best practical key recovery attack was on 820-round Trivium with complexity $2^{53.17}$. We put forward 820-, 825- and 830-round practical key-recovery attacks, in which there are $\mathbf{2^{80}\times 87.8\%}$, $\mathbf{2^{80}\times 83\%}$ and $\mathbf{2^{80}\times 65.7\%}$ keys that could be practically recovered, respectively, if we consider $\mathbf{2^{60}}$ as the upper bound for practical computational complexity. Besides, even for computers with computational power not exceeding $\mathbf{2^{52}}$ (resp. $\mathbf{2^{55}}$), we can still recover $\mathbf{58\%}$ (resp. $\mathbf{46.6\%}$) of the keys in the key space for 820 rounds (resp. 830 rounds). Our attacks have led 10 rounds more than the previous best practical attack.

The post-quantum digital signature scheme CRYSTALS-Dilithium has been recently selected by the NIST for standardization. Implementing CRYSTALS-Dilithium, and other post-quantum cryptography schemes, on embedded devices raises a new set of challenges, including ones related to performance in terms of speed and memory requirements, but also related to side-channel and fault injection attacks security. In this work, we investigated the latter and describe a differential fault attack on the randomized and deterministic versions of CRYSTALS-Dilithium. Notably, the attack requires a few instructions skips and is able to reduce the MLWE problem that Dilithium is based on to a smaller RLWE problem which can be practically solved with lattice reduction techniques. Accordingly, we demonstrated key recoveries using hints extracted on the secret keys from the same faulted signatures using the LWE with side-information framework introduced by Dachman-Soled et al. at CRYPTO’20. As a final contribution, we proposed algorithmic countermeasures against this attack and in particular showed that the second one can be parameterized to only induce a negligible overhead over the signature generation.

Picnic is a NIST PQC Round 3 Alternate signature candidate that builds upon symmetric primitives following the MPC-in-the-head paradigm. Recently, researchers have been exploring more secure/efficient signature schemes from conservative one-way functions based on AES, or new low complexity one-way functions like Rain (CCS 2022) and AIM (CCS 2023). The signature schemes based on Rain and AIM are currently the most efficient among MPC-in-the-head-based schemes, making them promising post-quantum digital signature candidates. However, the exact hardness of these new one-way functions deserves further study and scrutiny. This work presents algebraic attacks on RAIN and AIM for certain instances, where one-round Rain can be compromised in $2^{n/2}$ for security parameter $n\in \{128,192,256\}$, and two-round Rain can be broken in $2^{120.3}$, $2^{180.4}$, and $2^{243.1}$ encryptions, respectively. Additionally, we demonstrate an attack on AIM-III (which aims at 192-bit security) with a complexity of $2^{186.5}$ encryptions. These attacks exploit the algebraic structure of the power function over fields with characteristic 2, which provides potential insights into the algebraic structures of some symmetric primitives and thus might be of independent interest.

We introduce a novel side-channel-based reverse engineering technique capable of reconstructing a procedure solely from inputs, outputs, and traces of execution. Beyond generic restrictions, we do not assume any prior knowledge of the procedure or the chip it operates on. These restrictions confine our analysis to 8-bit RISC constant-time software implementations. Specifically, we demonstrate the feasibility of reconstructing a symmetric cryptographic cipher, even in scenarios where traces are sampled with information loss and noise, such as when measuring the power consumption of the chip.

In CRYPTO 2019, Gohr showed that well-trained neural networks could perform cryptanalytic distinguishing tasks superior to differential distribution table (DDT)-based distinguishers. This suggests that the differential-neural distinguisher (ND) may use additional information besides pure ciphertext differences. However, the explicit knowledge beyond differential distribution is still unclear. In this work, we provide explicit rules that can be used alongside DDTs to enhance the effectiveness of distinguishers compared to pure DDT-based distinguishers. These rules are based on strong correlations between bit values in right pairs of XOR-differential propagation through addition modulo $2^n$. Interestingly, they can be closely linked to the earlier study of the multi-bit constraints and the recent study of the fixed-key differential probability. In contrast, combining these rules does not improve the NDs' performance. This suggests that these rules or their equivalent form have already been exploited by NDs, highlighting the power of neural networks in cryptanalysis. In addition, we find that to enhance the differential-neural distinguisher's accuracy and the number of rounds, regulating the differential propagation is imperative. Introducing differences into the keys is typically believed to help eliminate differences in encryption states, resulting in stronger differential propagations. However, differential-neural attacks differ from traditional ones as they don't specify output differences or follow a single differential trail. This questions the usefulness of introducing differences in a key in differential-neural attacks and the resistance of Speck against such attacks in the related-key setting. This work shows that the power of differential-neural cryptanalysis in the related-key setting can exceed that in the single-key setting by successfully conducting a 14-round key recovery attack on Speck32/64.

The threat of physical side-channel attacks and their countermeasures is a widely researched field. Most physical side-channel attacks rely on the unavoidable influence of computation or storage on voltage or current fluctuations. Such data-dependent influence can be exploited by, for instance, power or electromagnetic analysis. In this work, we introduce a novel non-invasive physical side-channel attack, which exploits the data-dependent changes in the impedance of the chip. Our attack relies on the fact that the temporarily stored contents in registers alter the physical characteristics of the circuit, which results in changes in the die's impedance. To sense such impedance variations, we deploy a well-known RF/microwave method called scattering parameter analysis, in which we inject sine wave signals with high frequencies into the system's power distribution network (PDN) and measure the echo of the signals. We demonstrate that according to the content bits and physical location of a register, the reflected signal is modulated differently at various frequency points enabling the simultaneous and independent probing of individual registers. Such side-channel leakage violates the $t$-probing security model assumption used in masking, which is a prominent side-channel countermeasure. To validate our claims, we mount non-profiled and profiled impedance analysis attacks on hardware implementations of unprotected and high-order masked AES. We show that in the case of profiled attack, only a single trace is required to recover the secret key. Finally, we discuss how a specific class of hiding countermeasures might be effective against impedance leakage.

DeepCover is a secure authenticator circuit family developed by Analog Devices. It was designed to provide cryptographic functions, true random number generation, and EEPROM secure storage. DS28C36 is one of the DeepCover family, which is widely used in secure boot and secure download for IoT. It has been recently deployed in the Coldcard Mk4 hardware wallet as a second secure element to enhance its security. In this paper, we present for the first time, a detailed evaluation for the DS28C36 secure EEPROM against Laser Fault Injection (LFI). In the context of a black box approach, we prove by experimental results that the chip resists single fault attacks. In order to overcome this, we present the use of leakage detection such as Welch’s T-test to facilitate finding the correct moments for injecting successful faults, which is not common in Fault Injection (FI) as this method has been used only for Side-Channel Attacks (SCAs). By using this knowledge, we found two moments for injecting laser pulses to extract the protected EEPROM user pages with 99% success rate. The attack can be reproduced within a day. The presented attack negatively impacts the users of DS28C36 (including Coldcard Mk4).

Feistel network and its generalizations (GFN) are another important building blocks for constructing hash functions, e.g., Simpira v2, Areion, and the ISO standard Lesamnta-LW. The Meet-in-the-Middle (MitM) is a general paradigm to build preimage and collision attacks on hash functions, which has been automated in several papers. However, those automatic tools mostly focus on the hash function with Substitution-Permutation network (SPN) as building blocks, and only one for Feistel network by Schrottenloher and Stevens (at CRYPTO 2022). In this paper, we introduce a new automatic model for MitM attacks on Feistel networks by generalizing the traditional direct or indirect partial matching strategies and also Sasaki’s multi-round matching strategy. Besides, we find the equivalent transformations of Feistel and GFN can significantly simplify the MILP model. Based on our automatic model, we improve the preimage attacks on Feistel-SP-MMO, Simpira-2/-4-DM, Areion-256/-512-DM by 1-2 rounds or significantly reduce the complexities. Furthermore, we fill in the gap left by Schrottenloher and Stevens at CRYPTO 2022 on the large branch (b > 4) Simpira-b’s attack and propose the first 11-round attack on Simpira-6. Besides, we significantly improve the collision attack on the ISO standard hash Lesamnta-LW by increasing the attacked round number from previous 11 to ours 17 rounds.

At ASIACRYPT 2022, Benedikt, Fischlin, and Huppert proposed the quantum herding attacks on iterative hash functions for the first time. Their attack needs exponential quantum random access memory (qRAM), more precisely {$2^{0.43n}$} quantum accessible classical memory (QRACM). As the existence of large qRAM is questionable, Benedikt et al. leave an open question on building low-qRAM quantum herding attacks. In this paper, we answer this open question by building a quantum herding attack, where the time complexity is slightly increased from Benedikt et al.'s $2^{0.43n}$ to ours $2^{0.46n}$, but {it does not need qRAM anymore (abbreviated as no-qRAM)}. Besides, we also introduce various low-qRAM {or no-qRAM} quantum attacks on hash concatenation combiner, hash XOR combiner, Hash-Twice, and Zipper hash functions.

Over the last few years, deep learning is becoming the most trending topic for the classical cryptanalysis of block ciphers. Differential cryptanalysis is one of the primary and potent attacks on block ciphers. Here we apply deep learning techniques to model differential cryptanaly- sis more easily. In this paper, we report a generic tool called NDDT1, us- ing deep neural classifier that assists to find differential distinguishers for symmetric block ciphers with reduced round. We apply this approach for the differential cryptanalysis of ARX-based encryption schemes HIGHT, LEA, SPARX and SAND. To the best of our knowledge, this is the first deep learning-based distinguisher for the mentioned ciphers. The result shows that our deep learning based distinguishers work with high accuracy for 14-round HIGHT, 13-Round LEA, 11-round SPARX and 14-round SAND128. The relationship between the hamming weight of input difference of a neural distinguisher and the corresponding maxi- mum round number of the cipher has been justified through exhaustive experimentation. The lower bounds of data complexity for differential cryptanalysis have also been improved.

The use of random seeds to a deterministic random bit generator to generate uniform random sampling has been applied multiple times in post-quantum algorithms. The finalists Dilithium and Kyber use SHAKE and AES to generate the random sequence at multiple stages of the algorithm. Here we characterize one of the sampleing techniques available in Dilithium for a random sequence of length 256 with the help of the neutrosophic Boolean function. The idea of the neutrosophic Boolean function came from the theory of neutrosophy and it is useful to study any ternary distributions. We present the non-existence of neutrobalanced bent functions specifically with respect to the sampling named SampleInBall in Dilithium.

The Lattice Isomorphism Problem (LIP) is the computational task of recovering, assuming it exists, a orthogonal linear transformation sending one lattice to another. For cryptographic purposes, the case of the trivial lattice $\mathbb Z^n$ is of particular interest ($\mathbb Z$LIP). Heuristic analysis suggests that the BKZ algorithm with blocksize $\beta = n/2 + o(n)$ solves such instances (Ducas, Postlethwaite, Pulles, van Woerden, ASIACRYPT 2022). In this work, I propose a provable version of this statement, namely, that $\mathbb Z$LIP can indeed be solved by making polynomially many calls to a Shortest Vector Problem (SVP) oracle in dimension at most $n/2 + 1$.

Indistinguishability obfuscation (IO) is at the frontier of cryptography research for several years. LV16/Lin17 obfuscation schemes are famous progresses towards simplifying obfuscation mechanism. In fact, these two schemes only constructed two compact functional encryption (CFE) algorithms, while other things were taken to AJ15 IO frame or BV15 IO frame. That is, CFE algorithms are inserted into AJ15 IO frame or BV15 IO frame to form a complete IO scheme. The basic structure of two CFE algorithms can be described in the following way. The polynomial-time-computable Boolean function is transformed into a group of low-degree low-locality component functions by using randomized encoding, while some public combination of values of component functions is the value of original Boolean function. The encryptor uses constant-degree multilinear maps (rather than polynomial-degree multilinear maps) to encrypt independent variables of component functions. The decryptor uses zero-testing tool of multilinear maps to obtain values of component functions (rather than to obtain values of independent variables), and then uses public combination to obtain the value of original Boolean function. In this paper we restrict IO to be a real white box (RWB). Under such restriction we point out that LV16/Lin17 CFE algorithms being inserted into AJ15 IO frame are invalid. More detailedly, such insertion makes the adversary gradually learn the shape of the function, therefore the scheme is not secure. In other words, such scheme is not a real IO scheme, but rather a garbling scheme. It needs to be said that RWB restriction is reasonable, which means the essential contribution of IO for cryptography research.

HALFLOOP is a family of tweakable block ciphers that are used for encrypting automatic link establishment (ALE) messages in high-frequency radio, a technology commonly used by the military, other government agencies, and industries that require high robustness in long-distance communications. Recently, it was shown in [DDLS22] that the smallest version of the cipher, HALFLOOP-24, can be attacked within a practical time and memory complexity. However, in the real-word ALE setting, it turns out that this attack requires waiting more than 500 years to collect the necessary amount of plaintext-tweak-ciphertext pairs fulfilling the conditions of the attack. In this paper, we present real-world practical attacks against HALFLOOP-24 which are based on a probability-one differential distinguisher. In our attacks, we significantly reduce the data complexity to three differential pairs in the chosen-plaintext (CPA) setting which is optimal in the sense that even a brute force attack needs at least six plaintext-tweak-ciphertext pairs to uniquely identify the correct key. Considering the same ALE setting as [DDLS22], this translates to a reduction from 541 years to 2 hours worth of intercepted traffic. Besides, we provide the first, non generic, public cryptanalysis of HALFLOOP-48 and HALFLOOP-96. More precisely, we present Demirci-Selçuk meet-in-the-middle attacks against full-round HALFLOOP-48 and round-reduced HALFLOOP-96 to recover the complete master key in a CPA setting. However, unlike the attacks on HALFLOOP-24, our attacks on the larger versions are only theoretical. Moreover, for HALFLOOP-96 the known generic time-memory trade-off attack, based on a flawed tweak handling, remains the strongest attack vector. In conclusion, we iterate what was already stated in [DDLS22]: HALFLOOP does not provide adequate protection and should not be used.

We show that the scheme [Clust. Comput. 25(1): 451-468, 2022] fails to keep anonymity, not as claimed. The scheme simply acknowledges that user anonymity is equivalent to protecting the target user's identity against exposure, while its long-term pseudo-identity can be exposed. We want to clarify that the true anonymity means that an adversary cannot attribute different sessions to different target users, even though the adversary cannot recover the true identifier from the long-term pseudo-identifier. We also clarify some misunderstandings in the scheme.

LV16/Lin17 IO schemes are famous progresses towards simplifying obfuscation mechanism. In fact, these two schemes only constructed two compact functional encryption (CFE) algorithms, while other things were taken to the AJ15 IO frame or BV15 IO frame. CFE algorithms are inserted into the AJ15 IO frame or BV15 IO frame to form a complete IO scheme. We stated the invalidity of LV16/Lin17 IO schemes. More detailedly, under reasonable assumption “real white box (RWB)” LV16/Lin17 CFE algorithms being inserted into AJ15 IO frame are insecure. In this paper, we continue to state the invalidity of LV16/Lin17 IO schemes. The conclusion of this paper is that LV16/Lin17 CFE algorithms being inserted into BV15 IO frame are insecure. The reasoning of this paper is composed of the following three steps. First, when LV16/Lin17 CFE algorithms are inserted into secret constants. Second, when all secret random numbers are changed into the BV15 IO frame, all secret random numbers must be changed into secret constants, component functions in LV16/Lin17 CFE algorithms are cryptologic weak functions, and shapes of these component functions can be easily obtained by chosen values of independent variables. Finally, the shapes of these component functions include parameters of original function, therefore the IO scheme is insecure.

This research paper explores the vulnerabilities of the lightweight block cipher SPECK 32/64 through the application of differential analysis and deep learning techniques. The primary objectives of the study are to investigate the cipher’s weaknesses and to compare the effectiveness of ResNet as used by Aron Gohr at Crypto2019 and DenseNet . The methodology involves conducting an analysis of differential characteristics to identify potential weaknesses in the cipher’s structure. Experimental results and analysis demonstrate the efficacy of both approaches in compromising the security of SPECK 32/64.

In the Learning with Errors (LWE) problem we are given a matrix $A \in \mathbb{Z}_q^{N \times N}$ and a target vector $\vec{t} \in \mathbb{Z}_q^N$ such that there exists small-norm $\vec{s}, \vec{e}\in \mathbb{Z}_q^N$ satisfying $A\cdot\vec{s} = \vec{t} + \vec{e} \bmod q$. Modern cryptosystems often sample $\vec{s}, \vec{e}$ from narrow distributions that take integer values in a small range $[-\eta, \eta].$ Kyber and Dilithium both choose $\eta=2$ and $\eta=3$ using either a Centered Binomial distribution (Kyber), or a uniform distribution (Dilithium). In this work, we address the fundamental question how hard the enumeration of LWE secret keys for narrow distributions with $\eta \leq 3$ is. At Crypto 21, May proposed a representation-based algorithm for enumerating ternary keys, i.e. the case $\eta = 1$, with a fixed number of $\pm 1$ entries. In this work, we extend May's algorithm in several ways. First, we show how to deal with keys sampled from a probability distribution as in many modern systems like Kyber and Dilithium, rather than with keys having a fixed number of entries. Second, we generalize to larger values $\eta= 2, 3$, thereby achieving asymptotic key guess complexities that are not far off from lattice estimates. E.g. for Kyber's Centered Binomial distribution we achieve heuristic time/memory complexities of $\mathcal{O}(2^{0.36N})$ for $\eta=2$, and $\mathcal{O}(2^{0.37N})$ for $\eta=3$. For Dilithium's uniform distribution we achieve heuristic complexity $\mathcal{O}(2^{0.38N})$ for $\eta=2$. Let $S$ be the Shannon entropy of Kyber/Dilithium keys. Then our algorithms runs in time about ${S}^{\frac 16}$, which greatly improves over the standard combinatorial Meet-in-the-Middle attack with complexity ${S}^{\frac 1 2}$. Our results also compare well to current lattice asymptotics of $2^{0.29 \beta}$, where the lattice parameter $\beta$ is roughly of size $\frac 4 5 N$. Thus, our analysis supports that Kyber secret keys are indeed hard to enumerate. Yet, we find it remarkable that a purely combinatorial key search is almost competitive with highly evolved lattice sieving techniques.

We introduce a novel algebraic approach for attacking the McEliece cryptosystem which is currently at the $4$-th round of the NIST competition. The contributions of the article are twofold. (1) We present a new distinguisher on alternant and Goppa codes working in a much broader range of parameters than \cite{FGOPT11}. (2) With this approach we also provide a polynomial--time key recovery attack on alternant codes which are distinguishable with the distinguisher \cite{FGOPT11}. These results are obtained by introducing a subspace of matrices representing quadratic forms. Those are associated with quadratic relations for the component-wise product in the dual of the Goppa (or alternant) code of the cryptosystem. It turns out that this subspace of matrices contains matrices of unusually small rank in the case of alternant or Goppa codes ($2$ or $3$ depending on the field characteristic) revealing the secret polynomial structure of the code. MinRank solvers can then be used to recover the secret key of the scheme. We devise a dedicated algebraic modeling in characteristic $2$ where the Gröbner basis techniques to solve it can be analyzed. This computation behaves differently when applied to the matrix space associated with a random code rather than with a Goppa or an alternant code. This gives a distinguisher of the latter code families, which contrarily to the one proposed in \cite{FGOPT11} working only in a tiny parameter regime is now able to work for code rates above $\frac{2}{3}$. It applies to most of the instantiations of the McEliece cryptosystem in the literature. It coincides with the one of \cite{FGOPT11} when the latter can be applied (and is therefore of polynomial complexity in this case). However, its complexity increases significantly when \cite{FGOPT11} does not apply anymore, but stays subexponential as long as the co-dimension of the code is sublinear in the length (with an asymptotic exponent which is below those of all known key recovery or message attacks). For the concrete parameters of the McEliece NIST submission \cite{ABCCGLMMMNPPPSSSTW20}, its complexity is way too complex to threaten the cryptosystem, but is smaller than known key recovery attacks for most of the parameters of the submission. This subspace of quadratic forms can also be used in a different manner to give a polynomial time attack of the McEliece cryptosystem based on generic alternant codes or Goppa codes provided that these codes are distinguishable by the method of \cite{FGOPT11}, and in the Goppa case we need the additional assumption that its degree is less than $q-1$, where $q$ is the alphabet size of the code.

\ascon is the final winner of the lightweight cryptography standardization competition $(2018-2023)$. In this paper, we focus on preimage attacks against round-reduced \ascon. The preimage attack framework, utilizing the linear structure with the allocating model, was initially proposed by Guo \textit{et al.} at ASIACRYPT 2016 and subsequently improved by Li \textit{et al.} at EUROCRYPT 2019, demonstrating high effectiveness in breaking the preimage resistance of \keccak. In this paper, we extend this preimage attack framework to \ascon from two aspects. Firstly, we propose a linearize-and-guess approach by analyzing the algebraic properties of the \ascon permutation. As a result, the complexity of finding a preimage for 2-round \ascon-\xof with a 64-bit hash value can be significantly reduced from $2^{39}$ guesses to $2^{27.56}$ guesses. To support the effectiveness of our approach, we find an actual preimage of all ‘0’ hash in practical time. Secondly, we develop a SAT-based automatic preimage attack framework using the linearize-and-guess approach, which is efficient to search for the optimal structures exhaustively. Consequently, we present the best theoretical preimage attacks on 3-round and 4-round \ascon-\xof so far.

We show that the key agreement scheme [IEEE Trans. Veh. Technol. 71(4): 3470-3479, 2022] fails to keep user anonymity, not as claimed.

In this paper, we investigate the security of several recent MAC constructions with provable security beyond the birthday bound (called BBB MACs) in the quantum setting. On the one hand, we give periodic functions corresponding to targeted MACs (including PMACX, PMAC with parity, HPxHP, and HPxNP), and we can recover secret states using Simon algorithm, leading to forgery attacks with complexity $O(n)$. This implies our results realize an exponential speedup compared with the classical algorithm. Note that our attacks can even break some optimally secure MACs, such as mPMAC+-f, mPMAC+-p1, mPMAC+-p2, mLightMAC+-f, etc. On the other hand, we construct new hidden periodic functions based on SUM-ECBC-like MACs: SUM-ECBC, PolyMAC, GCM-SIV2, and 2K-ECBC$_{-}$Plus, where periods reveal the information of the secret key. Then, by applying Grover-meets-Simon algorithm to specially constructed functions, we can recover full keys with $O(2^{n/2}n)$ or $O(2^{m/2}n)$ quantum queries, where $n$ is the message block size and $m$ is the length of the key. Considering the previous best quantum attack, our key-recovery attacks achieve a quadratic speedup.

Generalized Feistel schemes (GFSs) are extremely important and extensively researched cryptographic schemes. In this paper, we investigate the security of Type-1 GFS in quantum circumstances. On the one hand, in the qCCA setting, we give a new quantum polynomial-time distinguisher on $(d^2-1)$-round Type-1 GFS with branches $d\geq3$, which extends the previous results by $(d-2)$ rounds. This leads to a more efficient analysis of type-1 GFS, that is, the complexity of some previous key-recovery attacks is reduced by a factor of $2^{\frac{(d-2)k}{2}}$, where $k$ is the key length of the internal round function. On the other hand, for CAST-256, which is a certain block cipher based on Type-1 GFS, we give a 17-round quantum distinguisher in the qCPA setting. Based on this, we construct an $r (r>17)$-round quantum key-recovery attack with complexity $O(2^{\frac{37(r-17)}{2}})$.

At Eurocrypt`22 Tang, Duong, Joux, Plantard, Qiao, and Susilo proposed a digital signature algorithm based on the hardness of the isomorphism problem of alternating trilinear forms. They propose three concrete parameters in dimensions $9$, $10$, and $11$ respectively. We give new heuristic algorithms that solve this problem more efficiently. With our new algorithms, the first parameter set can be broken in less than a day on a laptop. For the second parameter set, we show there is a $2^{-17}$ fraction of the public keys that can also be broken in less than a day. We do not break the third parameter set in practice, but we claim it falls short of the target security level of $128$ bits.

The lattice-based CRYSTALS-Dilithium signature scheme has been selected for standardization by the NIST. As part of the selection process, a large number of implementations for platforms like x86, ARM Cortex-M4, or – on the hardware side – Xilinx Artix-7 have been presented and discussed by experts. While software implementations have been subject to side-channel analysis with several attacks being published, an analysis of Dilithium hardware implementations and their peculiarities has not taken place. With this work, we aim to fill this gap, presenting an analysis of vulnerable operations and practically showing a successful profiled Simple Power Analysis (SPA) and a Correlation Power Analysis (CPA) on a recent hardware implementation by Beckwith et al. Our SPA attack requires 700 000 profiling traces and targets the first Number-Theoretic Transform (NTT) stage. After finishing profiling, we can identify pairs of coefficients with 1 101 traces. The full CPA attack finds secret coefficients with as low as 66 000 traces. In response, we present specific countermeasures and show that they effectively prevent both attacks.

Thus far, several papers reported concrete resource estimates of Shor's quantum algorithm for solving the elliptic curve discrete logarithm problem (ECDLP). In this paper, we study quantum FLT-based inversion algorithms over binary elliptic curves. There are two major algorithms proposed by Banegas et al. and Putranto et al., where the former and latter algorithms achieve fewer numbers of qubits and smaller depths of circuits, respectively. We propose two quantum FLT-based inversion algorithms that essentially outperform previous FLT-based algorithms and compare the performance for NIST curves of the degree $n$. Specifically, for all $n$, our first algorithm achieves fewer qubits than Putranto et al.'s one without sacrificing the number of Toffoli gates and the depth of circuits, while our second algorithm achieves smaller depths of circuits without sacrificing the number of qubits and Toffoli gates. For example, when $n = 571$, the number of qubits of our first algorithm is 74 \% of that of Putranto et al.'s one, while the depth of our second algorithm is 83 \% of that of Banegas et al.'s one. The improvements stem from the fact that FLT-based inversions can be performed with arbitrary sequences of addition chains for $n - 1$ although both Banegas et al. and Putranto et al. follow fixed sequences that were introduced by Itoh and Tsujii's classical FLT-based inversion. In particular, we analyze how several properties of addition chains, which do not affect the computational resources of classical FLT-based inversions, affect the computational resources of quantum FLT-based inversions and find appropriate sequences.

The Shortest Vector Problem (SVP) is at the foundation of lattice-based cryptography. The fastest known method to solve SVP in dimension $d$ is by lattice sieving, which runs in time $2^{td+o(d)}$ with $2^{md+o(d)}$ memory for constants $t,m \in \Theta(1)$. Searching reduced vectors in the sieve is a problem reduced to the configuration problem, \ie searching $k$ vectors satisfying given constraints over their scalar products. In this work, we present a framework for $k$-sieve algorithms: we filter the input list of lattice vectors using a code structure modified from [BDGL16] to get lists centred around $k$ codewords summing to the null-vector. Then, we solve a simpler instance of the configuration problem in the $k$ filtered lists. Based on this framework, we describe classical sieves for $k=3$ and $4$ that introduce new time-memory trade-offs. We also use the $k$-Lists algorithm [KMPM19] inside our framework, and this improves the time for $k=3$ and gives new trade-offs for $k=4$.