MIG-COW: Transferable Adversarial Attacks on Deepfake Detectors

# MIG-COW: Transferable Adversarial Attacks on Deepfake Detectors

https://gcdradar.vercel.app/incidents/mig-cow-transferable-adversarial-attacks-on-deepfake-detecto · AIR-2026-1473 · closed 2026-11-06

- Lane: Research
- Severity: 1 (Research / proof of concept)
- Fix status: Fix status unknown
- First reported: 2026-10-23
- AI role: Attack on AI
- Products: MIG-COW

## Summary

Researchers propose MIG-COW, an adversarial attack framework generating imperceptible inputs to bypass heterogeneous deepfake detectors. The method achieves high white-box attack success rates across various CNN and ViT models.

## What to do

No action recorded; read the sources below and the vendor's advisory.

## Latest updates

- 2026-10-23: MIG-COW: Transferable Adversarial Attacks on Deepfake Detectors via Gradient Decomposition (https://arxiv.org/abs/2610.30914)

## Sources

- 2026-10-23 arXiv: MIG-COW: Transferable Adversarial Attacks on Deepfake Detectors via Gradient Decomposition https://arxiv.org/abs/2610.30914

Verify against the vendor's advisory before changing production. Data CC BY-SA 4.0, https://gcdradar.vercel.app.