🤖 AI Summary
Unintentional leakage of sensitive information in visual multimedia (images/videos) poses significant privacy and security risks. Method: This paper systematically surveys and unifies adversarial and defensive techniques for visual data anonymization, proposing the first comprehensive taxonomy encompassing masking, inpainting, generative restoration, differential privacy, adversarial perturbations, and watermark-based provenance tracing. It formalizes threat models and establishes standardized robustness evaluation criteria. Contribution/Results: We construct a technical landscape synthesizing 120+ works, identifying critical bottlenecks—including poor generalizability and inadequate adaptability to dynamic scenes. Furthermore, we introduce a verifiable privacy-preserving research framework grounded in formal privacy guarantees, offering both theoretical foundations and practical guidelines for trustworthy visual data sharing.
📝 Abstract
The exploding rate of multimedia publishing in our networked society has magnified the risk of sensitive information leakage and misuse, pushing the need to secure data against possible exposure. Data sanitization -- the process of obfuscating or removing sensitive content related to the data -- helps to mitigate the severe impact of potential security and privacy risks. This paper presents a review of the mechanisms designed for protecting digital visual contents (i.e., images and videos), the attacks against the cited mechanisms, and possible countermeasures. The provided thorough systematization, alongside the discussed challenges and research directions, can pave the way to new research.
