Exploring the Feasibility of Low Cost Fault Injection Attacks on Sub-threshold Devices through an Example of a 65nm AES Implementation

The continuous scaling of VLSI technology and the aggressive use of low power strategies (such as subthreshold voltage) make it possible to implement standard cryptographic primitives within the very limited circuit and power budget of RFID devices. On the other hand, such cryptographic implementations raise concerns regarding their vulnerability to both active and passive side channel attacks. In particular, when focusing on RFID targeted designs, it is important to evaluate their resistance to low cost physical attacks. A common low cost fault injection attack is the one which is induced by insufficient supply voltage of the chip with the goal of causing setup time violations. This kind of fault attack relies on the possibility of gracefully degrading the performance of the chip. It is however, unclear whether this kind of low cost attack is feasible in the case of low voltage design since a reduction of the voltage may result in a catastrophic failure of the device rather than an isolated setup violation. Furthermore, the effect that process variations may have on the fault model used by the attacker and consequently the success probability of the attack, are unknown. In this paper, we investigate these issues by evaluating the resistance to low cost fault injection attacks of chips implementing the AES cipher that were manufactured using a 65nm low power library and operate at subthreshold voltage. We show that it is possible to successfully breach the security of a custom implementation of the AES cipher. Our experiments have taken into account the expected process variations through testing of multiple samples of the chip. To the best of our knowledge, this work is the first attempt to explore the resistance against low cost fault injection attacks on devices that operate at subthreshold voltage and are very susceptible to process variations.