New framework tests adversarial robustness of multigraph neural networks for financial crime analysis

Researchers have developed a new framework to test the robustness of multigraph neural networks (MGNNs) against adversarial attacks in financial crime analysis. The framework adapts projected gradient descent and projected randomized block coordinate descent methods to optimize structural perturbations in transaction networks. Experiments on synthetic and realistic datasets demonstrated that these learned attacks significantly reduce detection performance, while adversarial training effectively improves the robustness of MGNNs against such manipulations. AI

IMPACT This research highlights potential vulnerabilities in AI models used for financial crime detection and proposes methods for enhancing their security against sophisticated attacks.

RANK_REASON Academic paper detailing a new framework and experimental results. [lever_c_demoted from research: ic=1 ai=1.0]

Read on arXiv cs.LG →

• arXiv
• Financial Crime Analysis
• Hugging Face
• Multigraph Neural Networks
• Projected Gradient Descent
• Projected Randomized Block Coordinate Descent

• paper
• safety

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