Penetration testing on AI-generated impersonating using spoofing, voice cloning, and real-time deepfaking, 2025

AI-driven phishing, spoofing, and deepfake attacks pose escalating risks across email systems, educational platforms, and digital social spaces. Publicly available tools now enable adversaries to generate convincing impersonations, bypass conventional defenses, and exploit human trust through email spoofing, voice cloning, and real-time face swapping. These threats highlight the limitations of traditional security awareness training and automated filters when faced with adaptive, AI-powered exploits. This thesis develops a gamified cybersecurity training framework to address the educational and technical gaps in preparing users for such attacks. The framework combines interactive games with phishing and deepfake awareness exercises, improving both engagement and retention compared to conventional instruction. The proposed solution, CyberGLA (Gamified Learning Applications), integrates two complementary stages: (1) EmailKnight, an AI-assisted detection tool that analyzes headers, content, attachments, and metadata to identify spoofed or malicious emails, and (2) gamified training modules that adaptively deliver attack-specific education. Together, these reinforce echnical defenses with active, human-centered learning. Evaluation demonstrates improved detection of advanced phishing threats, measurable gains in user awareness, and stronger preparedness in both individual and institutional contexts. The research underscores the urgent need for adaptive security frameworks that integrate technical defenses with effective, engaging education to counter AI-enabled attacks.