Distance similarity measure based profiling attacks against masked chips

Abstract

The attack price of higher order side channel attacks (HOSCA) grows exponentially with the mask order. When the mask order is larger than 3, HOSCA can be inapplicable. In light of this, the distance similarity measure based profiling attack (DSMPA) is proposed. In DSMPA, the leakage property of masks can be characterised in a profiling scenario and the so called templates can be obtained. Then, DSM can be used to recover mask values with templates. With mask values recovered, the effect of masking disappears and one can use first-order side-channel attacks to recover the secret key used by a masked chip. The key-recovery efficiency of DSMPA is evaluated in real scenarios, and the advantage of DSMPA over HOSCA in terms of the key-recovery efficiency can be obvious under the success rate metric. Overall, the contribution of this work may provide evaluators a powerful tool to evaluate the security of masked chips.