1. 2

Today, resistance to physical defaults is a necessary criterion for masking schemes. In this context, the focus has long been on designing masking schemes guaranteeing security in the presence of glitches. Sadly, immunity against glitches increases latency as registers must stop the glitch propagation. Previous works could reduce the latency by removing register stages but only by impractically increasing the circuit area. Nevertheless, some relatively new attempts avoid glitches by applying DRP logic styles. Promising works in this area include LMDPL, SESYM - both presented at CHES - and Self-Timed Masking - presented at CARDIS - enabling to mask arbitrary circuits with only one cycle latency. However, even if glitches no longer occur, there are other physical defaults that may violate the security of a masked circuit. Imbalanced delay of dual rails is a known problem for the security of DRP logic styles such as WDDL but not covered in formal security models. In this work, we fill the gap by presenting the delay-extended probing security model, a generalization of the popular glitch-extended probing model, covering imbalanced delays. We emphasize the importance of such a model by a formal and practical security analysis of LMDPL, SESYM, and Self-Timed Masking. While we formally prove the delay-extended security of LMDPL and Self-Timed Masking, we show that SESYM fails to provide security under our defined security model what causes detectable leakage through experimental evaluations. Hence, as the message of this work, avoiding glitches in combination with d-probing security is not enough to guarantee physical security in practice.

Archive

  1.