Abstract
The discrete logarithm problem is a fundamental challenge in number theory with significant implications for cryptographic protocols. In this paper, we investigate the limitations of gradient-based methods for learning the parity bit of the discrete logarithm in finite cyclic groups of prime order. Our main result, supported by theoretical analysis and empirical verification, reveals the concentration of the gradient of the loss function around a fixed point, independent of the logarithm's base used. This concentration property leads to a restricted ability to learn the parity bit efficiently using gradient-based methods, irrespective of the complexity of the network architecture being trained. Our proof relies on Boas-Bellman inequality in inner product spaces and it involves establishing approximate orthogonality of discrete logarithm's parity bit functions through the spectral norm of certain matrices. Empirical experiments using a neural network-based approach further verify the limitations of gradient-based learning, demonstrating the decreasing success rate in predicting the parity bit as the group order increases.
Original language | English |
---|---|
Pages (from-to) | 1321-1336 |
Number of pages | 16 |
Journal | Proceedings of Machine Learning Research |
Volume | 222 |
Publication status | Published - 2023 |
Event | 15th Asian Conference on Machine Learning, ACML 2023 - Istanbul, Turkey Duration: Nov 11 2023 → Nov 14 2023 |
Keywords
- Cryptographic Protocols
- Discrete Logarithm
- Gradient-based Learning
ASJC Scopus subject areas
- Artificial Intelligence
- Software
- Control and Systems Engineering
- Statistics and Probability