Quantifying and Mitigating Cache Side Channel Leakage with Differential Set
summary
Introduces a new abstraction differential set and a tool DSA that both measures and removes cache side‑channel leakage by aligning sibling memory accesses, taking Cartesian products of local sets, and refining with past conditions to rule out impossible traces. DSA uses CBMC for pointer resolution, computes an upper bound then tightens it (example: naive product 18 → refined traces 4 → 2 bits leaked), reports ground‑truth leakage on benchmarks where CacheAudit overestimates, and produces mitigated code with 6.7% of leaking accesses hardened instead of touching whole arrays.
AI-written summary · about
chapters (16)
- 0:00Cache side channel research intro
- 1:03Cache channels leak data
- 2:08Modeling side channels
- 3:56Trace abstraction limits
- 4:38Differential set abstraction
- 6:04Global differential set
- 7:28Local differential sets
- 8:10Mitigating leakage
- 9:35Security verification
- 10:18Refining differential sets
- 11:00Branch alignment
- 12:26Quantification evaluation
- 13:07Leakage case studies
- 13:50Mitigation case studies
- 14:36Conclusion and future
- 14:56Q&A
related talks
15mSpecSafe: Detecting Cache Side Channels in a Speculative World
27mCacheD: Identifying Cache-Based Timing Channels in Production Software
15mI Know What You Said: Unveiling Hardware Cache Side-Channels in Local Large
13mExposing Cache Timing Side-Channel Leaks through Out-of-Order Symbolic Execution
25mHardening Confidential Federated Computations against Side-Channel Attacks
01h 01mTutorial: Side-Channel Attacks and Transient Execution Vulnerab...
28mStrong and Efficient Cache Side-Channel Protection...
20mA Formal Model of Cache Speculation Side-Channels
