Paper 2013/802 Original: eprint.iacr.org/2013/802 ↗

Cryptosystems Resilient to Both Continual Key Leakages and Leakages from Hash Functions

Abstract

Yoneyama et al. introduced Leaky Random Oracle Model (LROM for short) at ProvSec2008 in order to discuss security (or insecurity) of cryptographic schemes which use hash functions as building blocks when leakages from pairs of input and output of hash functions occur. This kind of leakages occurs due to various attacks caused by sloppy usage or implementation. Their results showed that this kind of leakages may threaten the security of some cryptographic schemes. However, an important fact is that such attacks would leak not only pairs of input and output of hash functions, but also the secret key. Therefore, LROM is rather limited in the sense that it considers leakages from pairs of input and output of hash functions alone, instead of taking into consideration other possible leakages from the secret key simultaneously. On the other hand, many other leakage models mainly concentrate on leakages from the secret key and ignore leakages from hash functions for a cryptographic scheme exploiting hash functions in these leakage models. Some examples show that the above drawbacks of LROM and other leakage models may cause insecurity of some schemes which are secure in the two kinds of leakage model. In this paper, we present an augmented model of both LROM and some leakage models, which both the secret key and pairs of input and output of hash functions can be leaked. Furthermore, the secret key can be leaked continually during the whole life cycle of a cryptographic scheme. Hence, our new model is more universal and stronger than LROM and some leakage models (e.g. only computation leaks model and bounded memory leakage model). As an application example, we also present a public key encryption scheme which is provably IND-CCA secure in our new model.