Open problems in technical ai governance

A Reuel, B Bucknall, S Casper, T Fist, L Soder… - arxiv preprint arxiv …, 2024 - arxiv.org
AI progress is creating a growing range of risks and opportunities, but it is often unclear how
they should be navigated. In many cases, the barriers and uncertainties faced are at least …

Zero-knowledge proofs of training for deep neural networks

K Abbaszadeh, C Pappas, J Katz… - Proceedings of the 2024 …, 2024 - dl.acm.org
A zero-knowledge proof of training (zkPoT) enables a party to prove that they have correctly
trained a committed model based on a committed dataset without revealing any additional …

Causes and effects of unanticipated numerical deviations in neural network inference frameworks

A Schlögl, N Hofer, R Böhme - Advances in Neural …, 2024 - proceedings.neurips.cc
Hardware-specific optimizations in machine learning (ML) frameworks can cause numerical
deviations of inference results. Quite surprisingly, despite using a fixed trained model and …

Unforgeability in stochastic gradient descent

T Baluta, I Nikolic, R Jain, D Aggarwal… - Proceedings of the 2023 …, 2023 - dl.acm.org
Stochastic Gradient Descent (SGD) is a popular training algorithm, a cornerstone of modern
machine learning systems. Several security applications benefit from determining if SGD …

Proof-of-learning with incentive security

Z Zhao, Z Fang, X Wang, X Chen, H Su, H **ao… - arxiv preprint arxiv …, 2024 - arxiv.org
Most concurrent blockchain systems rely heavily on the Proof-of-Work (PoW) or Proof-of-
Stake (PoS) mechanisms for decentralized consensus and security assurance. However, the …

Gradients look alike: Sensitivity is often overestimated in {DP-SGD}

A Thudi, H Jia, C Meehan, I Shumailov… - 33rd USENIX Security …, 2024 - usenix.org
Differentially private stochastic gradient descent (DP-SGD) is the canonical approach to
private deep learning. While the current privacy analysis of DP-SGD is known to be tight in …

Survey on Blockchain-Enhanced Machine Learning

O Ural, K Yoshigoe - IEEE Access, 2023 - ieeexplore.ieee.org
The convergence of blockchain and Machine Learning (ML) promises to reshape
technological innovation by enhancing security, efficiency, and transparency in ML systems …

Attesting distributional properties of training data for machine learning

V Duddu, A Das, N Khayata, H Yalame… - … on Research in …, 2024 - Springer
The success of machine learning (ML) has been accompanied by increased concerns about
its trustworthiness. Several jurisdictions are preparing ML regulatory frameworks. One such …

Zkdl: Efficient zero-knowledge proofs of deep learning training

H Sun, T Bai, J Li, H Zhang - IEEE Transactions on Information …, 2024 - ieeexplore.ieee.org
The recent advancements in deep learning have brought about significant changes in
various aspects of people's lives. Meanwhile, these rapid developments have raised …

Enhancing Security of Proof-of-Learning against Spoofing Attacks using Feature-Based Model Watermarking

O Ural, K Yoshigoe - IEEE Access, 2024 - ieeexplore.ieee.org
The rapid advancement of machine learning (ML) technologies necessitates robust security
frameworks to protect the integrity of ML model training processes. Proof-of-Learning (PoL) …