Embedded systems carry and process more and more
sensitive information in untrusted environments, where an attacker
can wiretap the external communication and also has
unlimited physical access to the device. Cryptography protects
systems against many of the threats and relies on the security of
the cryptographic keys inside the system.
Physical Unclonable Functions (PUFs) measure manufacturing
variations inside integrated circuits, for example FPGAs, to generate
a unique secret PUF response inside each device. Similar to
deriving a biometric pattern from human features, the individual
pattern inside an FPGA differs slightly from measurement to
measurement. From these measurements, the PUF response is
generated to derive a secure and reliable cryptographic key.
The Ring-Oscillator (RO) PUF is a popular PUF type because
of its high randomness and reliability. Frequencies of ROs are
compared pairwise to derive one secret bit. So far, the reliability
of RO PUFs was evaluated by counting bit flips in measured PUF
responses.
This work analyzes the distribution of frequency measurements
to derive the behavior of the PUF. Analyzing the frequency
distributions gives a more precise estimation of the PUF bit error
rates than measuring the bit errors after the comparison of two
oscillator frequencies.
The evaluation of publicly available real world empirical FPGA
data has shown that most error probabilities of RO PUF
responses are so low that they cannot be measured in feasible
time. For almost 200 evaluated FPGAs, more than 70% of the
PUF outputs on every FPGA have bit error probabilities under
10−20. We can even ensure this error probability for over 60%
of the PUF outputs after a practicable number of frequency
measurements with a confidence of 99.9%.
«
Embedded systems carry and process more and more
sensitive information in untrusted environments, where an attacker
can wiretap the external communication and also has
unlimited physical access to the device. Cryptography protects
systems against many of the threats and relies on the security of
the cryptographic keys inside the system.
Physical Unclonable Functions (PUFs) measure manufacturing
variations inside integrated circuits, for example FPGAs, to generate
a unique secret PUF resp...
»
Login
BibTeX