Trigger Value Selection in Clean-Label Malware Backdoors: An Effectiveness-Detectability Analysis
The study investigates how selecting specific values for trigger features impacts clean-label backdoor attacks against malware classifiers. Researchers tested various selection strategies like SHAP and quantile-based methods, finding that aggressive selectors boost attack success but can make the attack easier to detect by sanitization tools. Conversely, some selectors maintain effectiveness while reducing the likelihood of being caught.