Systematic Evaluation of Randomized Cache Designs against Cache Occupancy

Learn about a comprehensive security research presentation that systematically evaluates five randomized cache designs (CEASER, CEASER-S, MIRAGE, ScatterCache, and SassCache) against cache occupancy attacks, addressing a critical gap in current literature that primarily focuses on contention-based attacks while overlooking cache occupancy vulnerabilities. Discover how researchers from Max Planck Institute for Security and Privacy and Indian Institutes of Technology propose a uniform benchmarking strategy to fairly compare randomized cache designs across different replacement policies, revealing inadequacies in contemporary benchmarking approaches due to varying cache configurations and implementation assumptions. Explore the security evaluation methodology that examines three distinct threat models including covert channels, process fingerprinting, and AES key recovery attacks, with this work being the first to demonstrate complete AES key recovery on randomized cache designs using cache occupancy techniques. Understand the performance implications of different randomized cache architectures including randomized design, randomized-and-partitioned design, and pseudo-fully associative design approaches. Gain insights into why cache occupancy side-channels must be considered alongside traditional contention-based defenses when designing secure randomized cache systems, and examine the trade-offs between security and performance across various cache randomization strategies.