Invasive PUF Analysis

Abstract

In this work we consider the suitability of Phyiscaly Unclonable Functions (PUFs) for high-security applications. For PUFs to be considered secure in such scenarios they must be resilient to both semi-invasive and fully-invasive attacks. We introduce a new failure analysis technique for semi-invasive, single-trace, backside readout of logic states. We apply this technique to characterize the unique physical response of a memory-based PUF. With these results we identify several weakness in current PUF schemes. We extend current PUF definitions to be resilient against such attacks by requiring that PUFs be implemented in a serialized manner. Finally, we improve already existing PUF architectures to include these concepts.