Correlated quantization cuts communication in private federated RL
Original title: Communication-Efficient Private Federated Reinforcement Learning with Post-Quantum Authentication for Vehicular Networks
One-bit uploads with correlated private quantization reduce mean travel time by 14.3% while preserving privacy in vehicular networks.
The work studies private federated reinforcement learning for cooperative vehicular routing without centralizing trajectory data. Correlated private quantization using complementary shared randomness between paired vehicles reduces aggregation error while meeting local differential privacy. One-bit uploads shrink authenticated uplink size by 92.4% and raise mean verified updates per round. Under matched privacy the extension yields lower analytical error than independent quantization and resists vehicle-origin inference attacks.
- 92.4% reduction in authenticated uplink size
- 16.5% lower analytical aggregation mean squared error
- 14.3% reduction in mean travel time
- Inference attack accuracy drops to 22%
- 26.5 mean verified updates per round
- AI models
- reinforcement learning models
From the original article
1 min read at arxiv.orgFederated reinforcement learning supports cooperative vehicular routing without centralizing trajectory data, but model updates may reveal private information and incur communication overhead over intermittent vehicle-to-infrastructure links, particularly when post-quantum authentication is required.
By Hojat Allah Salehi · arxiv.org
A short excerpt; the article and its images belong to the publisher. Summary, key facts and assessment above are AI-written from the article and can be wrong.