The Rise of Tractable Circuits: From Cryptography to Continuous Generative Models
summary
Shows two concrete advances: compiled AES mixed-column into SDDs and multiplies those logic circuits with measured leakage models to do fast, (almost) exact inference — initial naive product of nine circuits blows up, but sparsified leakage distributions (zeroing the 1−ε tail and renormalizing) yields tractable multiplication and outperforms belief‑propagation SASCA on toy AES side‑channel benchmarks (success measured by top‑k ranking). Second, defines integral nodes via numerical quadrature to embed continuous latent variables into probabilistic circuits, giving nicer samples and new SOTA on several PC benchmarks.
AI-written summary · about
chapters (7)
- 0:00Introduction to circuits
- 3:33Properties of PC structure
- 7:16Inference in circuits
- 13:06Structured path prediction
- 17:15Side-channel cryptanalysis
- 27:30Future work in DNA storage
- 29:36Continuous generative models
related talks
12mLeakage models are a leaky abstraction (...)
25mThe Whole is Greater than the Sum of its Parts: Linear Garbling and Applications
37mFrontiers in Cryptography
40mPost-Quantum
01h 02mCryptographic Side-Channels on Embedded Devices
01h 00mA Brief History of Practical Garbled Circuit Optimizations
54mAttribute-based Encryption for Circuits
31mContention in Cryptography: Obfuscation, Leakage, UCE and the AI-DHI Assumptions