分野別セミナー

Recent research shows that many public-key or identity-based encryption schemes are vulnerable to side-channel attacks on the keys by the interaction of an adversary with a physical device. In order to tolerate the possible key leakage, leakage-resilient cryptography models a class of leakage output by allowing the adversary to specify a computable leakage function and to obtain the partial keys or, possibly, other internal states from the output of this function.
In this talk, we propose a leakage-resilient hidden-vector encryption scheme that supports the predicate operators such as conjunction, disjunction, comparison, range query, subset query etc. The proposed scheme is leakage-resilient attribute-hiding secure in the sense that the adversary not only can obtain the tokens of any non-match vector but also can learn a bounded amount of information on any key (non-match/match key).
We prove the security with a series of computationally indistinguishable games that uses the dual system encryption mechanism. We also analyze and discuss the performance of leakage bound parameters and leakage fraction for the practical security level. Finally, we give a scheme that achieves both attribute-hiding and payload-hiding, and also analyze the performance of schemes over larger alphabets.