Racing Gadget
Agile and Stealthy Cache Attacks - Streamlined Racers
BTech Project - EE491
Guides: Prof. Biswabandan Panda, Prof. Madhav Desai (Stand-in EE Department Advisor)
We explored speculative execution to speed covert-channels, building on prior works such as Streamline (ASPLOS ’21) and Racing Gadgets from Hacky Racers (ASPLOS ’23). The focus is on overcoming the measurement bottleneck in cache timing attacks by exploiting speculative execution and prefetcher-like mechanisms to improve bit-rate and reduce error-rate. The proposed approach achieves a bit transmission rate of 2.424 MB/s with an error-rate of 3.62%, making it about 15% faster than Streamline. This work demonstrates a speculation-based decoding mechanism and highlights challenges in synchronization and branch predictor mistraining.
Links: