# A New Benchmark and Model for Challenging Image Manipulation Detection

arXiv: https://arxiv.org/abs/2311.14218 · HTML: https://mlsanity.com/paper/2311.14218/

## Key facts

- Cited by 8 papers in the map (data through 14 Oct 2026).
- Builds on 16 of its 37 references that are in the map.
- First posted on arXiv 23 Nov 2023.

## Builds on (most cited first, 16 of 16)

- [Attention Is All You Need](https://mlsanity.com/paper/1706.03762/) (2017)
- [PyTorch: An Imperative Style, High-Performance Deep Learning Library](https://mlsanity.com/paper/1912.01703/) (2019)
- [Squeeze-and-Excitation Networks](https://mlsanity.com/paper/1709.01507/) (2017)
- [Rethinking Atrous Convolution for Semantic Image Segmentation](https://mlsanity.com/paper/1706.05587/) (2017)
- [Deep High-Resolution Representation Learning for Visual Recognition](https://mlsanity.com/paper/1908.07919/) (2019)
- [Multi-Scale Context Aggregation by Dilated Convolutions](https://mlsanity.com/paper/1511.07122/) (2015)
- [HigherHRNet: Scale-Aware Representation Learning for Bottom-Up Human Pose Estimation](https://mlsanity.com/paper/1908.10357/) (2019)
- [DeepFakes and Beyond: A Survey of Face Manipulation and Fake Detection](https://mlsanity.com/paper/2001.00179/) (2020)
- [Fighting Fake News: Image Splice Detection via Learned Self-Consistency](https://mlsanity.com/paper/1805.04096/) (2018)
- [Noiseprint: a CNN-based camera model fingerprint](https://mlsanity.com/paper/1808.08396/) (2018)
- [Image Manipulation Detection by Multi-View Multi-Scale Supervision](https://mlsanity.com/paper/2104.06832/) (2021)
- [SPAN: Spatial Pyramid Attention Network forImage Manipulation Localization](https://mlsanity.com/paper/2009.00726/) (2020)
- [DOLG: Single-Stage Image Retrieval with Deep Orthogonal Fusion of Local and Global Features](https://mlsanity.com/paper/2108.02927/) (2021)
- [Constrained R-CNN: A general image manipulation detection model](https://mlsanity.com/paper/1911.08217/) (2019)
- [A Full-Image Full-Resolution End-to-End-Trainable CNN Framework for Image Forgery Detection](https://mlsanity.com/paper/1909.06751/) (2019)
- [Comprint: Image Forgery Detection and Localization using Compression Fingerprints](https://mlsanity.com/paper/2210.02227/) (2022)

## Built on by (most cited first, 8 of 8)

- [REALEDIT: Reddit Edits As a Large-scale Empirical Dataset for Image Transformations](https://mlsanity.com/paper/2502.03629/) (2025)
- [AnimeDL-2M: Million-Scale AI-Generated Anime Image Detection and Localization in Diffusion Era](https://mlsanity.com/paper/2504.11015/) (2025)
- [Webly-Supervised Image Manipulation Localization via Category-Aware Auto-Annotation](https://mlsanity.com/paper/2508.20987/) (2025)
- [DEAL-300K: Diffusion-based Editing Area Localization with a 300K-Scale Dataset and Frequency-Prompted Baseline](https://mlsanity.com/paper/2511.23377/) (2025)
- [Training-Free In-Context Forensic Chain for Image Manipulation Detection and Localization](https://mlsanity.com/paper/2510.10111/) (2025)
- [EfficientIML: Efficient High-Resolution Image Manipulation Localization](https://mlsanity.com/paper/2509.08583/) (2025)
- [TruEye: Fine-Grained Detection of AI-Generated Human Subjects in Images](https://mlsanity.com/paper/2606.27505/) (2026)
- [BG-REAL: A Public Real-Data Anchored Benchmark for Background Manipulation Detection and Localization](https://mlsanity.com/paper/2607.26232/) (2026)

## Abstract (arXiv)

The ability to detect manipulation in multimedia data is vital in digital forensics. Existing Image Manipulation Detection (IMD) methods are mainly based on detecting anomalous features arisen from image editing or double compression artifacts. All existing IMD techniques encounter challenges when it comes to detecting small tampered regions from a large image. Moreover, compression-based IMD approaches face difficulties in cases of double compression of identical quality factors. To investigate the State-of-The-Art (SoTA) IMD methods in those challenging conditions, we introduce a new Challenging Image Manipulation Detection (CIMD) benchmark dataset, which consists of two subsets, for evaluating editing-based and compression-based IMD methods, respectively. The dataset images were manually taken and tampered with high-quality annotations. In addition, we propose a new two-branch network model based on HRNet that can better detect both the image-editing and compression artifacts in those challenging conditions. Extensive experiments on the CIMD benchmark show that our model significantly outperforms SoTA IMD methods on CIMD.