How Effective Are Publicly Accessible Deepfake Detection Tools? A Comparative Evaluation of Open-Source and Free-to-Use Platforms

Collection of arXiv papers (Mar 6, 2026) highlighting multiple practical security and privacy risks across ML, telecommunications, hardware side‑channels, and cryptographic systems. Key findings include: public deepfake detection tools show inverse tradeoffs (forensic tools high recall/low specificity; AI classifiers high specificity/low recall) and humans outperform automated tools; diffusion-based image editing can effectively erase robust invisible watermarks, degrading provenance signals; dataset‑distillation enables a new high-success, low-sample model‑hijacking attack (Osmosis Distortion