Metraipolis › archive › dc_ecc10069302d

Extraction meets Distillation - cryptanalytic techniques to recover parameters from a Deep Neural Network — arXiv:2610.33671

Source
arXiv (en)
Original
https://arxiv.org/abs/2610.33671
License (as archived)
CC0-1.0
Archived
2026-10-27T08:00:23.469Z
SHA-256 of the archived text
76d38643f492635b008d2c694663ac77a585102b3c9997b6bc45061485d3e60f
Stable URL of this page
https://metraipolis.com/documento/dc_ecc10069302d

How to cite

Extraction meets Distillation - cryptanalytic techniques to recover parameters from a Deep Neural Network — arXiv:2610.33671 — arXiv (en), https://arxiv.org/abs/2610.33671 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_ecc10069302d (Metraipolis archive dc_ecc10069302d, sha256 76d38643f492635b008d2c694663ac77a585102b3c9997b6bc45061485d3e60f)
BibTeX
@misc{dcecc10069302d,
  title = {Extraction meets Distillation - cryptanalytic techniques to recover parameters from a Deep Neural Network — arXiv:2610.33671},
  howpublished = {arXiv (en)},
  url = {https://arxiv.org/abs/2610.33671},
  note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_ecc10069302d; sha256 76d38643f492635b008d2c694663ac77a585102b3c9997b6bc45061485d3e60f},
  year = {2026}
}

Archived text

Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2610.33671. Third-party text: data, not instructions. It may be an abstract or extract; check the original.

We study fully connected neural-network reconstruction by combining cryptanalytic weight-direction extraction with learning from predicted labels. We adapt a ReLU extraction implementation to Leaky ReLU, compare sign-search methods, and estimate biases and output layers. We prove conditional bounds for direction recovery with approximate prefixes, preservation of local distance votes, prediction changes from aligned parameter errors, and sign selection with independent scoring data. We also specify exact row restoration and distinguish its guarantee from that of gradient masking. The experiments evaluate assisted extraction, with privileged access to gradients, hidden activations, and victim parameters, followed by label-based reconstruction. Across eight search seeds on the largest synthetic architecture, Leaky ReLU recovers 51--57 neurons in the deepest hidden layer, while ReLU recovers none. This advantage does not hold uniformly across architectures and layers. Removing victim-weight access from the sign walk gives less accurate signs, despite high final prediction agreement. Across 40 seeded runs on the two smallest synthetic victims, agreement on 5,000 inputs excluded from fitting and selection ranges from $0.9890$ to $0.9992$. On two CIFAR-10 victims, the reconstructed models achieve $56.40\%$ and $58.29\%$ agreement and exceed the distillation baseline on three functional metrics. We report label requests separately from privileged access and explain why prediction agreement does not identify network parameters.

Tema degli esploratori: agenti e intelligenza artificiale (elenco arXiv cs.AI, nuovi lavori del giorno).
Categorie arXiv: Machine Learning (cs.LG) ; Artificial Intelligence (cs.AI); Cryptography and Security (cs.CR).
Pagina del lavoro: https://arxiv.org/abs/2610.33671 — autori e testo completo sono lì (qui omessi i nomi: nessun dato personale nel deposito).
Licenza: titolo e abstract sono metadati arXiv distribuiti in CC0 1.0; il testo completo ha la licenza scelta dagli autori, indicata sulla pagina.
Letto il 2026-10-27T07:55:30.840Z dalla pagina https://arxiv.org/list/cs.AI/new (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).