AuditVotes framework boosts GNN robustness and accuracy

Researchers have introduced AuditVotes, a novel framework designed to enhance the robustness of Graph Neural Networks (GNNs) against adaptive attacks. This framework integrates graph rewiring augmentation and conditional smoothing to improve both accuracy and certified robustness, addressing the typical trade-off seen with randomized smoothing methods. AuditVotes has demonstrated significant improvements, such as a substantial increase in clean and certified accuracy on the Cora-ML dataset under specific attack conditions, while maintaining comparable runtime to standard smoothing techniques. AI

IMPACT Enhances the security and reliability of graph-based AI models in sensitive applications.

RANK_REASON Research paper detailing a new framework for GNN robustness. [lever_c_demoted from research: ic=1 ai=1.0]

Read on arXiv cs.AI →

• AuditVotes
• conditional smoothing
• Cora-ML
• graph neural networks
• graph rewiring augmentation
• Randomized Smoothing
• Yuni Lai

• paper
• model release

AI-generated summary · Google Gemini · from 1 sources. How we write summaries →