Improving Divide and Conquer Attacks against Cryptosystems by Better Error Detection / Correction Strategies

Divide and conquer attacks try to recover small portions of cryptographic keys one by one. Usually, a wrong guess makes subsequent ones useless. Hence possible errors should be detected and corrected as soon as possible. In this paper we introduce a new (generic) error detection and correction strategy. Its efficiency is demonstrated at various examples, namely at a power attack, two timing attacks against RSA implementations with and without Chinese Remainder Theorem, and a timing attack against the future AES (Rijndael). As the design of efficient countermeasures requires a good understanding of an attack's actual power, the possible improvement induced by sophisticated error detection and correction should not be neglected. Although divide and conquer attacks are typical for side-channel attacks, we would like to stress that they are not restricted to that field, as will be illustrated by Siegenthaler's attack.