Machine Learning Attack Series: Repudiation Threat and Auditing

Summary

This post examines the repudiation threat, where an attacker such as Mallory replaces the Husky AI model file with a backdoored version and then denies having done so. It shows how the Linux Audit Daemon (auditd) can record who accessed the model file and when, using auditctl rules tagged with a keyword and searched with ausearch, with logs pushed to centralized systems such as Splunk, the Elastic Stack, or Azure Sentinel.

Related items

• InfoRolling the cyber dice with open-source and open-weight AI modelsSimilar attack · CSO Online
• InfoAEGIS: Real-Time Latent-Space Backdoor Detection for Dependable and Secure Small Language Model InferenceSimilar attack · OpenAlex (peer-reviewed AI security)
• InfoHFL-Deflect: Defence framework against adaptive coordinated poisoning attacks in hierarchical federated learningSimilar attack · Elsevier Security Journals
• InfoSecurity Threat Detection and Defense Theory of Intelligent Operation and Maintenance System for Rail Transit Based on Adversarial Machine LearningSimilar attack · OpenAlex (peer-reviewed AI security)
• InfoSwitchable backdoor attack in vision transformers via progressive dual-token injection from shallow to deep layersSimilar attack · OpenAlex (peer-reviewed AI security)