NDSS 2025 – Automated Mass Malware Factory

A blog highlights an NDSS 2025 paper on an automated approach to mass-produce Android malware by combining piggybacking with adversarial examples. The method extracts a malicious rider, generates tailored adversarial perturbations, and selects benign carriers to create malware that evades detectors, achieving reported ASRs of 88.3% on ML models (Drebin, MaMaDroid) and 76%/92% on Microsoft/Kingsoft engines. The work also discusses potential defenses.