After more than 20 year's research and development, side-channel attacks are constantly posing serious threats to various computing systems.
When targeting crypto-implementations to retrieve the secret, side-channel attacks utilize the peculiarity of the specific implementations, and achieve much better efficiency than brute force attacks and traditional cryptanalysis which attacks the weakness of the cryptographic algorithms themselves.
Typical side channels include power consumption, electromagnetic emanation, and execution time.
With inherent correlation between these side-channel information and the secret, statistic analysis can be employed to find the secret.
However, there are still many challenges presented for side-channel research driven by two trends: new ciphers and emerging computing platforms.
New ciphers or variants are being developed to provide higher level of security or get tailored to different applications.For example, XTS-AES is a security-hardened mode of AES for storage systems, which increases the algorithm complexity and hides more system-dependent parameters to users~(attackers). Meanwhile, we see more emerging computing platforms, for general purpose computing or specific algorithm acceleration. Graphic Processing Unit~(GPU) has been used to run a range of cryptographic algorithms for higher performance. However, the security of GPU when processing sensitive data, especially the highly relevant side-channel vulnerabilities, has received little attention and is vastly unexplored. Yet GPU differs from other computing platforms distinctly in terms of the hardware structure and software programming model, making side-channel attacks on GPU much more challenging.
In this dissertation, I propose several novel side-channel attacks, targeting new ciphers including XTS-AES and ECC and also popular accelerators - GPUs.
Some of our vulnerabilities analysis and security evaluation are first of its kind, and we anticipate them to pave the way for mitigations and lead to more active side-channel research.
- Professor Yunsi Fei (Advisor)
- Professor David Kaeli
- Professor Aatmesh Shrivastava
- Dr. Lei Poo