Securing Embedded Hardware
summary
FLAM‑PUF builds an LFSR+arbiter+feedback PUF, trades 64 cycles initialization for ML resistance, tests 10^6 CRPs and reports attacker prediction ≈50% while saving ~62% area versus prior designs. CaPUF composes 16 lightweight PL‑PUFs in a 4×4 cascade on VC707 boards, uses per‑PU frequency control and shows ML accuracy at coin‑flip after 65,000 challenges. Configurable Poisoning PUF splits challenges, maps to LPN problem for provable hardness and uses server‑side error correction; BLAST detects HLS Trojans by control/data‑trace equivalence checking.
AI-written summary · about
chapters (11)
- 0:00Start
- 1:11Wendy Chong
- 7:54Uniformity Uniqueness and Reliability
- 19:41Evaluate the Costs
- 20:33Machine Learning Modeling Attacks against Paths
- 28:24Predictability
- 35:33Learning Parity with Noise Problem
- 39:00Performance of Empirical Deep Learning Attacks
- 49:46Condition of Execution and Data Transformation
- 51:52Battery Exertion Attack
- 1:00:08Conclusions
related talks
21mWhat Is a PUF and Why Might You Need One?
44mTamper Proofing Commodity Hardware & other Applications
20mEnabling Low-Cost Secure Computing on Untrusted In-Memory Architectures
41mFault Attacks & Side Channels
20mEPF: Evil Packet Filter
59mNationally Critical Infrastructure needs Secure Hardware
48mCapturing 0Day Exploits With Perfectly Placed Hardware Traps
42mCPU Fuzzing: Automatic Discovery of...