Explore a comprehensive analysis of real-world adversarial promotional porn images (APPIs) used in underground advertising in this 19-minute conference talk presented at the 2019 IEEE Symposium on Security & Privacy. Delve into the strategies employed by cybercriminals to evade explicit content detection while preserving sexual appeal, despite observable distortions. Learn about Male`na, a novel deep learning-based methodology developed to understand these adversarial images and the underground business behind them. Discover findings from over 4,000 APPIs identified among 4,042,690 images crawled from popular social media, including evasion techniques against popular explicit content detectors and the ecosystem of illicit promotions. Gain insights into cybercriminal tactics, compromised account usage, large-scale APPI campaigns, and attempts to steal images for advertising purposes. Understand the importance of research on real-world adversarial learning and its potential for mitigating associated threats in online security and machine learning.
Overview
Syllabus
Stealthy Porn: Understanding Real World Adversarial Images for Illicit Online Promotion
Taught by
IEEE Symposium on Security and Privacy