ActivityForensics: A Comprehensive Benchmark for Localizing Manipulated Activity in Videos

2026-04-04

AI Summary

This work addresses the limitations of existing video forgery detection benchmarks, which predominantly focus on appearance-level manipulations and struggle to detect activity-level forgeries that alter human behaviors to distort event semantics. To bridge this gap, the study formally defines and constructs ActivityForensics—a large-scale benchmark for activity-level video tampering localization—comprising over 6,000 visually coherent forged clips. The authors further introduce TADiff, a diffusion-based feature regularization method designed to uncover subtle temporal artifacts indicative of manipulation. Comprehensive evaluations under cross-domain and open-world protocols systematically assess state-of-the-art approaches. ActivityForensics substantially advances the detection of highly deceptive activity-level forgeries and provides a robust foundation of data, methodology, and evaluation standards for future research in video forensics.

Abstract

Temporal forgery localization aims to temporally identify manipulated segments in videos. Most existing benchmarks focus on appearance-level forgeries, such as face swapping and object removal. However, recent advances in video generation have driven the emergence of activity-level forgeries that modify human actions to distort event semantics, resulting in highly deceptive forgeries that critically undermine media authenticity and public trust. To overcome this issue, we introduce ActivityForensics, the first large-scale benchmark for localizing manipulated activity in videos. It contains over 6K forged video segments that are seamlessly blended into the video context, rendering high visual consistency that makes them almost indistinguishable from authentic content to the human eye. We further propose Temporal Artifact Diffuser (TADiff), a simple yet effective baseline that exposes artifact cues through a diffusion-based feature regularizer. Based on ActivityForensics, we introduce comprehensive evaluation protocols covering intra-domain, cross-domain, and open-world settings, and benchmark a wide range of state-of-the-art forgery localizers to facilitate future research. The dataset and code are available at https://activityforensics.github.io.

Problem

• temporal forgery localization
• activity-level forgeries
• video manipulation
• media authenticity
• forged video detection

Innovation

• activity-level forgery
• temporal forgery localization
• diffusion-based regularizer
• video forensics benchmark
• TADiff