Leakage auditing at frontier scale
Establish methods for quantifying residual leakage from shared federated-learning weights under realistic multi-site conditions and for tracing that leakage to a responsible participant.
References
Best practices presume a threat model, but quantifying residual leakage from shared weights under realistic multi-site conditions is unvalidated at frontier scale, as is tracing leakage to a responsible participant.
— From Pilots to Production: Lessons in Cross-Institutional Federated Training and Artificial Intelligence for Science
(2609.39803 - Kotevska et al., 30 Sep 2026) in Section 4, “What Remains Unsolved, and a Shared Agenda”
Detecting passive inference robustly remains open.
— Quantization-Tolerant Gradient Auditing for Active Inference-Attack Detection in Federated Learning
(2610.13311 - Kuehne-Schlinkert, 9 Oct 2026) in Conclusion, Section 7
Model updates in on-device fine-tuning and federated learning may introduce additional privacy risks that remain for future study.
— BiShield-TEE: On the (In-)Security of Unilateral Weight Obfuscation in On-Device TEE-Shielded LLM Partition
(2610.21565 - Wang et al., 15 Oct 2026) in Section Discussion, Limitations