Attacked Test Set
Participants submit perturbed versions of the provided deepfake test set to induce misclassification by detectors.
Adversarial Attacks on Deepfake Detectors: Generated Media in Real-World Scenarios
The AADD 2026 Challenge is a grand challenge at ACM Multimedia 2026 focused on adversarial attacks against deepfake detectors under realistic, real-world conditions, including JPEG/JPEG-AI compression and social media-like processing. It builds on the 2025 edition by evaluating detector robustness using high- and low-quality generated media, aiming to advance forensic tools critical for digital trust and misinformation prevention.
Full paper
June 25, 2026
AoE
Conference timeline
Full paperKey deadline
June 25, 2026 · AoE
Paper fit
A strong submission should clearly identify its contribution and evaluate it appropriately.
Participants submit perturbed versions of the provided deepfake test set to induce misclassification by detectors.
A 1–2 page abstract describing methodology, motivation, and key contributions, required for evaluation.
Invited only to the top 3 teams; a detailed paper undergoes review and may be published in ACM Proceedings.
Compiled from the official call for papers. The organizers’ pages remain authoritative.
Last verified September 9, 2026