Unlocking Faster Encryption: Optimizing Symmetric Cryptography on GPUs 🚀

COSIC seminar – Optimization of Symmetric Cryptography Algorithms on GPUs – Cihangir Tezcan (Middle East Technical University)

Parallel computing power of GPUs can be used to optimize symmetric key encryption algorithms and these efficient implementations can be used to obtain fast encryption, perform brute force attacks on short keys, and verify theoretical results in practice. We can implement these algorithms generally in three ways: naïve, table-based, and bitsliced. In our recent works, we optimized 3DES, AES, ASCON, CRYPTO1, DES, and PRESENT algorithms for GPUs. By implementing CRYPTO1 in a bitsliced manner, we reduced the time required for offline attacks to clone Mifare Classic cards from months to hours. Our table-based implementation of AES-128 achieves 878.6 Gbps in counter mode on an RTX 2070 Super GPU which is 4.087 Gbps per Watt. Our table based optimizations provides 3.87 billion keys searches for DES and 3DES and 1.89 billion key searches per second for PRESENT. This result shows that 20 million RTX 3070 GPUs can capture an 80-bit PRESENT key in a year. Finally, our bitsliced implementation of ASCON allowed us to find the best distinguishers that we could not obtain theoretically by known methods and showed that theoretically obtained best distinguishers have better probabilities in practice. In this talk I am going to explain our optimizations and results.