Cybersecurity Expert Highlights AI-Driven Phishing Threats and Need for Advanced Detection

The presentation by Michel E. Parent, a cybersecurity and AI expert, focuses on the evolving threat of AI-driven phishing attacks and the need for resilient detection methods. He highlights that 59% of organizations worldwide were hit by ransomware in the past year, with roughly half of these attacks originating from phishing. AI tools now enable attackers to mass-produce highly persuasive, personalized phishing emails with minimal errors, increasing real-world click rates from 0.28% to 0.84% in the last 12 months. Traditional detection methods fail due to model drift—where classifiers trained on outdated data lose effectiveness—and human fatigue, as users struggle to identify sophisticated manipulation tactics like urgency, authority appeals, or scarcity. Parent proposes a dual approach combining static features (e.g., URL analysis with 20 structural indicators) and dynamic content scoring (e.g., 45 behavioral psychology-based features) to improve detection. He emphasizes explainability in classifiers to build user trust and combat alert fatigue, advocating for retraining models regularly to address drift. The talk references studies from Reuters, Harvard, and Microsoft, including a 2025 NetSkope report on rising click rates and an AI-obfuscated PDF payload case.