edfas.org 13 ELECTRONIC DEV ICE FA I LURE ANALYSIS | VOLUME 24 NO . 4 devices have become a major concern. Figure 1 exemplifies how counterfeit or illegal FPGAs within the supply chain can end up undetected in our computing systems. Amemometer is suggested to overcome this problem. The memometer is a low-overhead, inexpensive, adaptable hardware metering (fingerprinting) methodology leveraging memory physically unclonable functions (PUFs). Historically, memory PUFs have not been applied to contemporary FPGAs becausemost of themcomewith manufacturingmemorypreset startupvalues. Theauthors have overcome this issue by inventing a newmemory PUF using cross-coupled look-up tables (LUTs) that imitate the SRAM PUF behavior, thus providing unique start-up values (SUVs) used for fingerprinting each FPGA. These fingerprints are further used in identification and authentication throughout the supply chain. HARDWARE METERING Hardware metering helps in identifying authorship of an IC or intellectual property (IP) after fabrication by uniquely locking/tagging each IC that is manufactured under the samemask.[4] Hardwaremetering is further classified as passive and active metering.[4] Passive metering is used to tag each IC with an unclonable unique identifier. This identifier is further used in recognizing genuine ICs from the overbuilt/counterfeit ICs. Whereas in active metering, in addition to tracking passively, it can also help with enabling/disabling IC functionality and controlling/ preventing the ICs from further infiltrating the supply chain.[4] This passive metering methodology can be used to create unique unclonable fingerprints and use them to interrogate ICs within the supply chain. The authors are also leveraging themethodology to activelymeter, which is briefly described later. PHYSICALLY UNCLONABLE FUNCTIONS Identification and authentication are critical to secure any electronic system. Em- bedding a unique key can only help identify an IC, but in order to authenticate, a secret key must be embedded onto the IC itself.[5] These secret keys are either stored in nonvolatile memory (NVM) or battery-backed external volatile memory. Bothmethods not only add additional overhead but are also extremely vulnerable to attackers. A simple side-channel attack[6] can reveal a lot about the IC and allow for the secret key to be stolen, which can be further used in creating clones of those ICs. To overcome this issue, a new authentication mechanism—physically unclonable functions (PUFs)—was invented. PUFs are extremely hard-to-forge, unique to every IC ever manufactured, non-programmed, and low-overhead.[5,7] The basic idea behind a PUF is that each IC exhibits a unique process variation characteristic profile that can be leveraged to create unclonable fingerprints. Even when two ICs are functionally same, the underlying microscopic process variation characteristics are slightly different. When a challenge (input) Ci is applied to a section of an IC, the underlying unique process variation profile in that section exhibits a unique response (output) RCi. [7] Uniqueness and reproducibility are the two metrics used toanalyze thequalityof PUF fingerprints. Uniqueness ismeasured using inter-chip hamming distance (HD), and reliability or reproducibility is measured using intra-chip HD.[8] Inter-chip HD is the average HD measured between the responses when the same challenge is applied to two different ICs. Ideally, it should be 50%, which means half of the bits from these two fingerprints must be different. This measurement can also be used to analyze two fingerprints obtained from different sections of the same IC. Intra-chipHD is the average HDmeasured between the responseswhen the same challenge is applied at different times. Ideally, this should be 0%, which means that each fingerprint must be reproducible or repeatable over time. Anexampleof this illustration is shown inFig. 2. Aprogramming file (*.bit) is used as a challenge, applied on different FPGAs, and the response fingerprints are recorded. These responses are used to analyze the uniqueness of these fingerprints. Similarly, a challenge is applied to the same FPGAmultiple different times and the responses are used to investigate the repeatability measure of a given fingerprint. Fig. 2 (a) Uniqueness (inter-chipHD), (b) repeatability/reproducibility (intrachip HD) of PUF challenge-response pairs. (a) (b)
RkJQdWJsaXNoZXIy MTMyMzg5NA==