Object-Based Video Forgery Detection via Dual-Stream Networks

Abstract

The object-based video forgery detection aims to expose tampered regions from video sequences without any codec information. However, existing methods mainly focus on manually selected features and models for a specific task, either splicing or copy-move, while the general representation ability of deep learning models and the correlation of different forensic features have not been fully explored. In this paper, we propose a dual-stream framework to jointly discover and integrate effective features for object-based video forgery detection. First, two different types of branches are employed to extract discriminative features. Then, after the dualstream feature fusion, a Conditional Random Field (CRF) layer is utilized to further refine segmentation results. Finally, we consider temporal consistency by incorporating the video tracking strategy. Extensive experiments on four datasets show that the proposed method achieves competitive performance against the state-of-the-art methods.

Spatio-temporal co-attention fusion network for video splicing localization

“…• Video splicing localization network: OVFD. 2 The test results provided in their literature are cited, as publicly available source code is unavailable. • General video forensic network: VideoFact, 3 which we replicate using the author-provided code and datasets.…”

“…In comparison to images, the blind detection of video splicing is challenging due to the extra motion disturbance and higher data dimensions. 2 As pointed out by the prior work, 3 video compression inevitably weakens the forensic traces, resulting in the failure of image forensic techniques. As a result, it is significant to develop specialized forensic techniques for video splicing localization.…”

“…The UVL, 16 a unified end-to-end learning network, integrates spatial and frequency information to extract common features of tampering operations. Due to the absence of large-scale video forgery datasets, image datasets are used by Jin et al 2,17 to train video forensic networks. Video object tracking and segmentation are applied to temporally refine the preliminary localization maps within single frames.…”

“…Misuse of such video forgeries incurs serious implications on legal matters, public opinion, and intellectual property rights. In comparison to images, the blind detection of video splicing is challenging due to the extra motion disturbance and higher data dimensions 2 . As pointed out by the prior work, 3 video compression inevitably weakens the forensic traces, resulting in the failure of image forensic techniques.…”

Image Inpainting Forgery Detection: A Review

“…In [98] the authors are extending the work [99], by adding a multivariate process to enhance the detection phase and increasing the searching speedness. In [100] suggested using several Resnet networks (one for each type of forgery) plus the last one for following the forged area. As a basis for detecting inpainting, they are using a filter similar as in [87].…”