The increasing integration of AI and distributed systems — whether in the form of large-scale model training, edge inference, or autonomous multi-agent coordination—raises urgent challenges around reliability, dependability, and system-level trustworthiness. Meanwhile, AI itself offers promising tools for improving fault detection, anomaly response, and predictive resilience in distributed infrastructures.
Byzantine faults are particularly concerning in these settings. Their impact spans corrupted gradients in distributed training, faulty sensor readings in IoT infrastructure, and malicious nodes in edge/federated environments. Addressing Byzantine behavior is not only vital for ensuring robust distributed AI, but also for using AI effectively in monitoring and enhancing the dependability of broader system stacks.
The BFT-AI workshop, co-located with SRDS 2025, will bring together researchers working at the intersection of reliable distributed systems and artificial intelligence, with a dual focus:
- Designing AI systems (training or inference) that are resilient to Byzantine faults and adversarial environments.
- Leveraging AI-based techniques to detect, diagnose, and mitigate Byzantine behavior in complex distributed infrastructures.
Topics of Interest
- Fault-Tolerant AI: Resilient training and inference under Byzantine failures
- AI for Systems Resilience: Detecting and mitigating system faults using ML
- Theory: Convergence guarantees and models under adversarial conditions
- Infrastructure: Tools, datasets, reproducibility
Join us in shaping the future of the intersection between Byzantine fault tolerance and AI.
Important Dates
(All deadlines are at 23:59 AOE)
| Event | Date |
|---|---|
| Paper Submission | July 4, 2025 |
| Author Notification | July 25, 2025 |
| Camera-ready Deadline | July 31, 2025 |
| Workshop Date | September 29, 2025 |
Contacts
For further information, please contact us at bftaiws@gmail.com.