Attacked Test Set
Participants submit perturbed versions of the provided deepfake test images, following specified formats and guidelines.
Multimedia/CORE Unranked
AADD 2026
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, specifically targeting robustness under real-world conditions such as JPEG/JPEG-AI compression and social media processing. It builds on the 2025 edition by evaluating deepfake detection systems under more realistic and challenging scenarios to advance the development of reliable forensic tools for digital integrity.
Key deadlines
Verified September 9, 2026Full paper
June 25, 2026
AoE
Conference timeline
Submission and decisions
Full paperKey deadline
June 25, 2026 · AoE
Paper fit
Contribution paths
A strong submission should clearly identify its contribution and evaluate it appropriately.
Abstract Paper
A 1–2 page abstract describing the methodology, motivation, and key contributions of the proposed attack approach.
Full-Length Paper (Top 3 Teams Only)
Invited top 3 teams submit detailed full papers for review; accepted papers will be included in ACM Multimedia proceedings.
Research areas in scope
01
Adversarial Attacks on Deepfake Detectors
Adversarial perturbations on deepfake imagesRobustness under JPEG/JPEG-AI compressionSocial media-like post-processing (resizing, re-encoding)Detection evasion using GANs and diffusion modelsReal-world deployment scenarios for deepfake forensics
Policies worth checking twice
- Each team may submit only one attacked test set; only the last submission before the deadline is considered.
- Submissions must include original and perturbed image pairs; incomplete pairs are not evaluated.
- The challenge requires participation in ACM Multimedia 2026 as an on-site event; remote presentations are not allowed.
- Only the top 3 teams are eligible to submit full-length papers for review and potential inclusion in the ACM proceedings.
- Teams must submit an abstract paper along with their attacked test set to be eligible for evaluation.
Official sources
Compiled from the official call for papers. The organizers’ pages remain authoritative.
Last verified September 9, 2026