2026-10-03 –, STAGE 2 (Ghost in the Shellcode) Language: English
Deepfake detection using 3D CNNs traditionally focuses on identifying synthetic facial manipulations in video. This work inverts that lens. I leverage temporal‑spatial 3D CNN features originally designed to spot fake videos and apply them to a completely different domain: OSINT‑based profiling. By treating metadata as a volumetric signal across related individuals, the same convolutional filters that detect frame‑to‑frame anomalies can reconstruct familial linkages from public or leaked databases. This methodology enhances traditional OSINT by adding a predictive layer. Instead of manually building family trees, the 3D CNN model learns kinship patterns and flags potential person matches with high confidence. The result is a hybrid attack that combines AI forensics and open source intelligence to identify a target individual from a distant relative’s spit sample. I demonstrate a proof of concept using synthetic profiles and real OSINT sources. The talk shows how a deepfake detection technique becomes a privacy‑breaking weapon for person traceability at scale.
With over eight years of hands on experience in offensive security and vulnerability discovery, I specialize in data driven threat hunting across complex product ecosystems. Currently completing a Master of Science in Software Engineering with a focus on deepfake detection using 3D CNNs, I bridge the gap between academic research and advanced exploitation techniques. I have authored three technical volumes, including AI For Red Team Operation and Practical Application Security, while maintaining a global rank of 6 on Hackthebox. My work focuses on architecting resilient defense systems that integrate runtime application self protection and virtual patching to proactively decrease the time to pwn and neutralize emerging threats. By leveraging multi cloud telemetry and offensive research, I have identified over 140 vulnerabilities for major global vendors, ensuring product integrity against sophisticated adversaries.