Tag
This paper proposes GraphRP, a proactive defense framework using model reprogramming to protect GNNs from model extraction attacks, with a structure-aware gating mechanism that preserves benign utility while degrading adversarial queries.
Proposes a Test-Time Adaptive (TTA) composition framework for MLaaS in IoT environments, using a TTA-aware composability model and service-level adaptation to adjust services during inference, reducing computational time compared to traditional adaptive approaches.