When Hackers Build Their Own AI: Phishing, Recon & Deepfakes | Viktorya Markosyan скачать в хорошем качестве

When Hackers Build Their Own AI: Phishing, Recon & Deepfakes | Viktorya Markosyan

In her first-ever conference talk, Viktorya Markosyan uses HAL 9000 from Stanley Kubrick's 2001: A Space Odyssey as a lens for exploring the growing threat of AI-powered cyberattacks — a system designed to help that gradually turns against the people it was meant to serve. Viktorya covers three major areas where threat actors are already using AI effectively: Phishing: Since ChatGPT launched, AI-generated phishing emails have increased 1,000%. Human-crafted emails had a 12% click-through rate; AI-personalized emails now achieve 54%. Tools like FraudGPT (launched March 2023) lower the skill barrier dramatically, enabling even novice attackers to run convincing campaigns at massive scale. Reconnaissance: AI can ingest public documentation, API specs, Swagger files, blog posts, and error messages to correlate assets, rank attack targets by value, and generate attack path recommendations — in hours rather than days. Deepfakes: The most alarming section. Viktorya details the 2024 Hong Kong case where attackers used a deepfake video call impersonating a CFO and colleagues to convince a victim to wire $25 million across 15 transfers. She also discusses synthetic identity creation and AI-powered job applicant fraud. The talk closes with a sobering thought: AI doesn't question intent. It just does what it's told — forever. #AIHacking #Deepfakes #Phishing #CyberSecurity #ThreatIntelligence #APISECCON