by Michelle Zimmerman ’21
As image forgery technology becomes more and more pervasive, it is also increasingly crucial to be able to differentiate between real and fake images. Legal, political, and privacy reasons have driven the field of media forensics research to develop photo authentication methods. This summary paper will outline the most recent progress in the field, specifically the most popular algorithms that have been used to devise methods of image manipulation detection.
Although progress in deep learning is advancing faster than ever, there are significant steps that have been made in image forgery independent of deep learning. Non-deep learning algorithms are widely used to split an image and extract its feature vectors or sort those vectors, and continue to produce some of the most cutting-edge techniques today. Two main methods of comparison when searching for image forgery are block- and point-based methods. Both are used to help identify copy-move as well as splicing attacks by searching the image and comparing it against itself. The key to block-based methods is that the suspicious photo is split up into identical or irregular blocks which are then analyzed. By contrast, point-based methods extract key points of objects and their feature vectors, and this information is used to find tampering in the image. There are also hybrid methods which combine block and point-based algorithms, which conclusively achieves the same thing.
Furthermore, Support Vector Machines (SVMs) are supervised learning classifiers, often used along with the previously mentioned comparison methods or other algorithms to sort data and identify if and where an image has been tampered with.
Another important algorithm is Discrete Cosine Transform (DCT), which is a mechanism for organizing and extracting data. DCT breaks the image down into spectral sub-bands of different frequencies, containing different information. Researchers can then take advantage of this arrangement to find the data they need for their proposed method of image forgery detection. Sometimes, this method is also combined with some of the previously mentioned algorithms, such as SVMs.
Convolutional Neural Networks (CNNs) and deep learning algorithms in general are also very popular among image forgery detection research. This may be because using them can help easily create more comprehensive detection methods, opposed to past studies which can only distinguish specific types of manipulation, copy-move, splicing, etc.
In conclusion, although the technology for detecting fake images has come very far in recent years, so have algorithms that create these manipulated photos. In order to stay a step ahead, it’s crucial for research to continue into not only the mentioned methods but also other, currently more experimental ones. However, for those who do not work in the field, it is important to stay vigilant online when viewing and sharing images, gifs. or videos. The way this fake information spreads is through people, so if everyone makes an effort to be conscious about what they interact with, it can also make a large difference.
Tian-Tsong Ng and Shih-Fu Chang, “A Model for Image Splicing”, IEEE International Conference on Image Processing (ICIP), Singapore, October 2004.
Duc-Tien Dang-Nguyen, Cecilia Pasquini, Valentina Conotter, and Giulia Boato, “RAISE: a raw images dataset for digital image forensics” in MMSys ’15 Proceedings of the 6th ACM Multimedia Systems Conference, Pages 219-224, 2015.
E. Ardizzone, A. Bruno, and G. Mazzola, “Copy-Move Forgery Detection by Matching Triangles of Keypoints”, IEEE Transaction on Information Forensics and Security, vol. 10, no. 10, 2015.
I. Amerini, L. Ballan, R. Caldelli, A. Del Bimbo, and G. Serra, “A SIFT-based forensic method for copy-move attack detection and transformation recovery”, IEEE Transactions on Information Forensics and Security, vol. 6, no. 3, pp. 1099-1110, 2011.
V. Christlein, C. Riess, J. Jordan, C. Riess, and E. Angelopoulou: “An Evaluation of Popular Copy-Move Forgery Detection Approaches”, IEEE Transactions on Information Forensics and Security, vol. 7, no. 6, pp. 1841-1854, 2012.
D. Tralic, I. Zupancic, S. Grgic, and M. Grgic, “CoMoFoD – New Database for Copy-Move Forgery Detection,” in Proceedings of 55th International Symposium ELMAR, pp. 49–54, 2013.