New framework certifies ML model robustness against backdoor poisoning attacks

Researchers have developed a new framework for certifying the robustness of machine learning models against backdoor poisoning attacks. This framework utilizes $f$-divergences to handle complex adversary constraints, including the number of corrupted training examples and perturbation magnitudes. It offers a general approach that can recover previous certification methods as special cases and has been used to certify robustness against $\ell_p$-constrained poisoning adversaries with various noise distributions. AI

IMPACT Enhances the security of machine learning models against adversarial attacks, potentially increasing trust in their deployment.

RANK_REASON The item is a research paper published on arXiv detailing a new framework for machine learning model robustness. [lever_c_demoted from research: ic=1 ai=1.0]

Read on arXiv cs.LG →

• A Certified Robustness Framework for Backdoor Poisoning Attacks using $f$-Divergences
• alphaXiv
• arXiv
• CatalyzeX
• DagsHub
• Gaussian function
• Gotit.pub
• Hugging Face
• IArxiv
• Laplace
• machine learning
• Randomized Smoothing
• ScienceCast
• uniform

• paper
• safety

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