Unsupervised fusion for forgery localization exploiting background information

Ferrara, Pasquale; Fontani, Marco; Bianchi, Tiziano; Rosa, Alessia De; Piva, Alessandro; Barni, Mauro

doi:10.1109/icmew.2015.7169770

Cited by 13 publications

(7 citation statements)

References 18 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…In fact, to attribute the correct importance to a detector score, one should take into account its overall reliability, the level of confidence associated with the specific decision and its compatibility with other detectors. Experiments in [250], [251] show that the DST approach, with reasonable fusion rules, outperforms consistently all individual tools and also abstract-level fusion, thanks to the use of richer information. It even outperforms machine-learning fusion in the absence of a training set well-aligned with the test set, a recurrent situation in multimedia forensics.…”

Section: Fusionmentioning

confidence: 93%

“…Their results clearly show that detectors based on fusion dominate all individual base detectors in terms of accuracy, and the gain is more significant when operating at the abstract level. In [250], [251] fusion is addressed in a more systematic way, relying on the Dempster-Shafer theory (DST) of evidence [252] to meaningfully combine multiple tools at the measurement level. The DST provides a methodology to include concepts like uncertainty, reliability, and compatibility in the decision process.…”

Section: Fusionmentioning

confidence: 99%

See 1 more Smart Citation

Media Forensics and DeepFakes: an overview

Verdoliva¹

2020

Preprint

View full text Add to dashboard Cite

With the rapid progress of recent years, techniques that generate and manipulate multimedia content can now guarantee a very advanced level of realism. The boundary between real and synthetic media has become very thin. On the one hand, this opens the door to a series of exciting applications in different fields such as creative arts, advertising, film production, video games. On the other hand, it poses enormous security threats. Software packages freely available on the web allow any individual, without special skills, to create very realistic fake images and videos. So-called deepfakes can be used to manipulate public opinion during elections, commit fraud, discredit or blackmail people. Potential abuses are limited only by human imagination. Therefore, there is an urgent need for automated tools capable of detecting false multimedia content and avoiding the spread of dangerous false information. This review paper aims to present an analysis of the methods for visual media integrity verification, that is, the detection of manipulated images and videos. Special emphasis will be placed on the emerging phenomenon of deepfakes and, from the point of view of the forensic analyst, on modern data-driven forensic methods. The analysis will help to highlight the limits of current forensic tools, the most relevant issues, the upcoming challenges, and suggest future directions for research.

show abstract

Section: Fusionmentioning

confidence: 93%

Section: Fusionmentioning

confidence: 99%

Media Forensics and DeepFakes: an overview

Verdoliva¹

2020

Preprint

View full text Add to dashboard Cite

show abstract

“…Research has shown that many editing operations, such as resizing [2] or contrast enhancement [3], will leave unique traces behind. Many forensic algorithms have been developed to detect or identify editing operations [4][5][6][7][8][9][10][11][12][13][14][15]. In recent years, convolutional neural networks (CNNs) have been widely used by researchers to perform forensic tasks such as image tampering detection [9,[16][17][18] and source identification [19][20][21].…”

Section: Introductionmentioning

confidence: 99%

The Effect of Class Definitions on the Transferability of Adversarial Attacks Against Forensic CNNs

Zhao¹

2021

Preprint

View full text Add to dashboard Cite

In recent years, convolutional neural networks (CNNs) have been widely used by researchers to perform forensic tasks such as image tampering detection. At the same time, adversarial attacks have been developed that are capable of fooling CNN-based classifiers. Understanding the transferability of adversarial attacks, i.e. an attack's ability to attack a different CNN than the one it was trained against, has important implications for designing CNNs that are resistant to attacks. While attacks on object recognition CNNs are believed to be transferrable, recent work by Barni et al. has shown that attacks on forensic CNNs have difficulty transferring to other CNN architectures or CNNs trained using different datasets. In this paper, we demonstrate that adversarial attacks on forensic CNNs are even less transferrable than previously thought -even between virtually identical CNN architectures! We show that several common adversarial attacks against CNNs trained to identify image manipulation fail to transfer to CNNs whose only difference is in the class definitions (i.e. the same CNN architectures trained using the same data). We note that all formulations of class definitions contain the "unaltered" class. This has important implications for the future design of forensic CNNs that are robust to adversarial and anti-forensic attacks.

show abstract

“…There has been accordingly an increasing research in forgery image localization [1], [2], [3], [4], [8], [9], [10], [16], [17]. The general idea in forgery image detection is to find the local inconsistencies which is caused by the presence of traces left during tampering processes in the investigated image.…”

Section: Introductionmentioning

confidence: 99%

“…They proposed an energy-minimization approach using Markov Random Field to model the prior knowledge about the tampering maps. In [16], [17] the authors proposed a decision fusion framework for the image forensics scenario based on Dempster-Shafer Theory (DST) avoiding the necessity of assigning prior probabilities which would be difficult to estimate. However these methods only considered JPEG compression artifacts.…”

Section: Introductionmentioning

confidence: 99%

Effective images splicing detection based on decision fusion

Retraint

2019

2019 IEEE International Symposium on Signal Processing and Information Technology (ISSPIT)

View full text Add to dashboard Cite

Localizing the tampered regions in forgery images is an important and challenging problem in forensics applications. Though there have been an extensive studies on image forgery localization over past decade, each method still has its own limitations. Therefore, it is promising to fuse different forensic approaches in order to obtain better localization performance. In this paper, we propose a framework to aggregate the decision maps of two forensic approaches: Photo Response Non-Uniformity (PRNU) based approach and statistical features based approach using Dempster-Shafer Theory. PRNU noise can be considered as a camera fingerprint thereby being used effectively to localize tampering images. However, the most challenging limitation of this approach is its false identifications on textured, saturated and dark regions. By combining with the statistical feature based approach, we can decrease this false alarm rate on saturated and dark regions. The extensive experimental results demonstrate that the proposed method significantly outperforms the single PRNU based approach.

show abstract

Unsupervised fusion for forgery localization exploiting background information

Cited by 13 publications

References 18 publications

Media Forensics and DeepFakes: an overview

Media Forensics and DeepFakes: an overview

The Effect of Class Definitions on the Transferability of Adversarial Attacks Against Forensic CNNs

Effective images splicing detection based on decision fusion

Contact Info

Product

Resources

About