Research On Privacy Preserving For Perceptual Image Hashing In Social Network

The arrival of Web2.0 era provides an opportunity to the development and flourish of social network. And with the popularity of social network, people are sharing and publishing more and more information. Meanwhile there are a lot of privacy and security problems today which call for new privacy protection technologies. In recent years, the leakage and abuse of privacy information make people worry about the privacy issue of social network furtherly. This dissertation focuses on the most common information carriers in social network-images, and researches on the privacy preserving methods for digital image application.There are risks of copying, tampering and forgery when using large scale image resources in social network. Perceptual hashing plays an important role in digital forensics, pirate detection and source tracking of the massive networking information. Aiming at current academics pay less attention to privacy and security of the perceptual hashing, this dissertation proposes and designs a set of security protocols for privacy preserving digital image hashing. Specific encryption algorithms and hashing algorithms are designed on the basis of protocols. Experimental results prove the feasibilities of the protocols in social networks, providing privacy for the whole process of network applications. The main research contents include:Firstly, aiming at the problem that the traditional two-side architecture of perceptual hashing application is not suitable for distributed environment of social network, we propose a four-side hashing architecture and two basic definitions. The Quartet structure (image provider or owner, trusted forensics platform, trusted forensics platform center, and certification authority) ensures the separation of responsibilities, and thus provide a security computing framework for the protocols and algorithms of perceptual hashing in network environments. Two basic definitions are respectively "exchangeable hashing and encryption technology" and "distance preserving feature encryption algorithm". The former used to extract hash, makes the hash extracted from encrypted image equivalent to the encryption of hash extracted from the original image; the latter used to compare hash, makes the distance of encrypted hash consistent with the plaintext hash. Theoretical and experimental results have proved that the proposed basic definitions guarantee the feasibility of the architecture in network environment.Secondly, On the basis of the basic structure and the basic definitions, we design a kind of security enhanced hashing protocols based on the symmetric encryption and implement specific encryption algorithm and hashing algorithm based on the protocols. The security-enhanced protocols use the session key of user and computing platform to encrypt image. The hash extracted from the encrypted image has the same function with the traditional perceptal hash in authentication and identification. These protocols can be used in a network environment with the precondition that the computing platform is trustworthy. Experimental results prove that they can meet the social network which has high requirement for computational efficiency but general requirement for privacy preserving.Finally, On the basis of the basic structure and the basic definitions, we design a kind of security enhanced hashing protocols based on the public encryption and also implement specific algorithm based on the protocols. The security-enhanced protocols use user’s public key certificate and homomorphic encryption algorithm to encrypt image, while the computing platform is unable to decrypt the image. These protocols can be used in a network environment where the computing platform is untrustworthy. Experimental results prove that they can meet the social network which has high requirement for privacy preserving but general requirement for computational efficiency.