Description
Research about side-channel attacks on microcontrollers and other simpler devices is widespread, while
there is less research about side-channel attacks on more complex devices, in particular for attacks on
public-key cryptography. It is not clear which considerations need to be made when transferring an already
demonstrated attack to another device architecture. This thesis researches the influence that the
device architecture has on electromagnetic (EM) side-channel leakage of public-key cryptographic implementations
and what considerations need to be made when transferring an attack to a different device
architecture. This is done by measuring the EM side-channel leakage of four cryptographic implementations
each corresponding to one of four leakage models on 32-bit Armv7 and 64-bit Armv8 architecture
on an Arm Cortex-A72 processor. The four analyzed leakage models are the observation of control flow,
the observation of a single bit being parsed, the distinction between a full and a zero mask in a conditional
swap operation, and the hamming weight leakage model. The measured cryptographic implementations
are Mbed TLS RSA, OpenSSL RSA, OpenSSL elliptic-curve cryptography, and the reference implementation
of ML-KEM (formerly Kyber). The influence of the two architectures is analyzed by comparing the
execution time, the leakage at a single point in time measured by the Welch T-Test and the signal-to-noise
ratio, the frequencies in the EM trace, and the visibility of patterns in the trace. Using these criteria, the
transferability of an attack between architectures is evaluated, which can be used to motivate countermeasures
against side-channel attacks even on architectures that do not have a proven attack directly against
them.
|
Thesis topic in co-operation with the Fraunhofer Institute for Applied and Integrated Security AISEC, Garching