Random number generation is a key function of today's secure devices. Commonly used for key generation, random number streams are more and more frequently used as the anchor of trust of several countermeasures such as masking. True Random Number Generators (TRNGs) thus become a relevant entry point for attacks that aim at lowering the security of integrated systems. Within this context, this paper investigates the robustness of TRNGs based on Ring Oscillators (focusing on the delay chain TRNG) against pulsed electromagnetic fault injection. Indeed, weaknesses in generating random bits for masking scheme degenerate the Side Channel resistance. Finally by exploiting fault results on delay chain TRNG some general guidelines to harden them are derived.

Criterion principles Results Conclusion Why binding EM analysis to injection ? EMFI combinatory complexity: Time efficiency a : Analysis map → one day for three executables. Injection map (fixed parameters) → three days for one executable. a timing are relative to our setup Aim: Ease and fasten EMFI security characterisation → (X,Y) position. EMFISC 2017 3 / 19 Criterion principles Results Conclusion EMFI hotspots definition Designing the criterion

No abstract