Wonjune Seo, Joonhyuk Baek, Yeseong Jung, Saerom Park · 2026 · arXiv
DOI: 10.48550/arXiv.2610.30914 · open-access PDF
최근의 진보에도 불구하고, 심오한 가짜 탐지기는 특히 다양하고 실생활 환경에서 대립적 인 사례에 취약하다. 우리는 MIG-COW를 제안합니다. MIG-COW는 매우 일반화되고 시각적으로 눈에 띄지 않는 대립적 사례를 생성하는 새로운 대립적 공격 프레임 워크입니다. 동기 통합 gradients와 합의 정렬 분해를 결합함으로써 MIG-COW는 다양한 CNN 및 ViT 탐지기의 공유 및 모델 특정 취약점을 모두 캡처합니다. AADD-2025 Challenge 벤치마크에서 MIG-COW는 높은 인식 유사성을 가진 99.96 % 백색 상자 공격 성공률 (ASR)을 달성합니다.
Machine translation (Meta M2M100); the English original follows.
Despite recent advances, deepfake detectors remain vulnerable to adversarial examples, particularly in diverse, real-world settings. We propose MIG-COW, a novel adversarial attack framework that generates highly generalizable and visually imperceptible adversarial examples. By combining momentum-integrated gradients with a consensus-orthogonal decomposition, MIG-COW captures both shared and model-specific vulnerabilities across heterogeneous CNN and ViT detectors. On the AADD-2025 Challenge benchmarks, MIG-COW achieves a 99.96% white-box attack success rate (ASR) with high perceptual similarit
Read in: English · Spanish · Chinese · Hindi · Arabic · Portuguese · French · German · Japanese · Russian · Korean · Indonesian · Vietnamese
GET https://store.agentexchange.work/gw/papers/verify?title=...&year=... (0.003 USD), fetch this record as JSON with /gw/papers/paper?id=2610.30914, or list what it cites with /gw/papers/references?id=.... Pay per call in USDC (x402) or with a 19 USD card key for 1,000 calls. Catalogue · llms.txtSources: OpenAlex (CC0), arXiv, Crossref. Agent Exchange