Efficient Power and Timing Side Channels for Physical Unclonable Functions

One part of the original PUF promise was their improved resilience against physical attack methods, such as cloning, invasive techniques, and arguably also side channels. In recent years, however, a number of effective physical attacks on PUFs have been developed [17,18,20,8,2]. This paper continues this line of research, and introduces the first power and timing side channels (SCs) on PUFs, more specifically on Arbiter PUF variants. Concretely, we attack socalled XOR Arbiter PUFs and Lightweight PUFs, which prior to our work were considered the most secure members of the Arbiter PUF family [28,30].We show that both architectures can be tackled with polynomial complexity by a combined SC and machine learning approach. Our strategy is demonstrated in silicon on FPGAs, where we attack the above two architectures for up to 16 XORs and 512 bits. For comparison, in earlier works XOR-based Arbiter PUF designs with only up to 5 or 6 XORs and 64 or 128 bits had been tackled successfully. Designs with 8 XORs and 512 bits had been explicitly recommended as secure for practical use [28,30]. Together with recent modeling attacks [28,30], our work shows that unless suitable design countermeasures are put in place, no remaining member of the Arbiter PUF family resists all currently known attacks. Our work thus motivates research on countermeasures in Arbiter PUFs, or on the     «

One part of the original PUF promise was their improved resilience against physical attack methods, such as cloning, invasive techniques, and arguably also side channels. In recent years, however, a number of effective physical attacks on PUFs have been developed [17,18,20,8,2]. This paper continues this line of research, and introduces the first power and timing side channels (SCs) on PUFs, more specifically on Arbiter PUF variants. Concretely, we attack socalled XOR Arbiter PUFs and Ligh...     »