A Robust Open-Set Multi-Instance Learning for Defending Adversarial Attacks in Digital Image

In recent times, digital image forensics is gaining increased attention in multimedia forensics owing to the widespread scam alertness. Several forensic methods have been studied to establish the integrity of digital images by disclosing manipulation fingerprints. Anti-forensic (AF) attacks on manipulated images, particularly deep learning-based adversarial attacks using generative adversarial network (GAN), have been successfully applied to delude forensic methods. Consequently, an efficacious, efficient, and robust counter-AF (CAF) method is required to secure the integrity of digital images. In this study, we propose a robust open-set multi-instance learning approach for exposing GAN-based AF on manipulated images by introducing additional GAN-based operations. First, we generate multiple real instances from real images using multiple additional generators. Then we train an embedding network collaboratively with multiple real instances in an open-set fashion. During training, the embedding network learns only real images and has no prior knowledge regarding AF images. In the testing phase, real and AF images are processed for detection. The proposed open-set CAF method can effectively detect AF images and is more robust against transferable updating.

I. Introduction

Social platforms, including Facebook, Twitter, and YouTube, facilitate the sharing of images and videos on important occasions with friends and relatives. These images and videos can be manipulated for malicious use, such as spreading fake news with fake photos and demanding individual favors. Consequently, feigning digital images [1] for scamming has become a significant threat in multimedia forensics. Diverse image forgery techniques, which include compression, filtering, contrast enhancement, and deepfake technology, are used to alter the image content or generate a completely new image for malicious intentions. In addition to forgery techniques, various efficient and effective image forensic tools have been developed to detect feigned images based on handcrafted features and deep learning. However, the practice of anti-forensic (AF) attacks on manipulated images can easily misguide forensic methods. Consequently, we analyzed effects of AF attacks on manipulated images, and then proposed a robust counter-AF (CAF) method based on open-set multi-instance learning.