Design and Analysis of Multimedia Cryptosystems

多媒體加密算法的設計與密碼分析

Student thesis: Doctoral Thesis

View graph of relations

Author(s)

  • Yu ZHANG

Related Research Unit(s)

Detail(s)

Awarding Institution
Supervisors/Advisors
Award date23 Aug 2016

Abstract

The rapid increase of the popularity of the internet and multimedia communications has resulted in the fast development of information exchange and consumer electronics applications in recent years. However, it has also led to an increase in the demand of real-time and secure transmission of huge data sets. The easiest way to cope with this is to consider the multimedia streams as standard bit streams and apply traditional cryptographic approaches like Advanced Encryption Standard (AES) in proper modes of operation. In fact, the desire for cryptosystems to be more efficient with specifical designs for multimedia streams has drawn increasing attention from scientific communities in the past decade.
Regarding different design philosophies, two different approaches, direct encryption (DE) and selective encryption (SE), can be identified in the literature. Direct encryption refers to a model that directly encrypts multimedia data by specialized encryption algorithms utilizing the intrinsic features of multimedia data. While selective encryption, for the purpose of reducing the computational complexity and maintaining the semantics of the encrypted elements, refers to a model that incorporates ciphers into the source coding procedure and then selectively encrypts part of the data set. This thesis is mostly concerned with the cryptanalysis of some existing works and the design of some new algorithms in both directions.
The first part of the thesis begins with the cryptanalysis of some recently proposed image cryptosystems. Though all the considered ciphers are based on the structure of substitution-permutation network (SPN), which is a well known architecture used in block cipher algorithms such AES, their respective operations range from chaotic iterations, quantum walks, and the Chinese reminder theorem (CRT) to some traditional cryptographical functions like Exclusive OR and modulo addition. However, based on some newly discovered properties of CRT and a difference equation of modulo addition, our analysis indicates that their respective equivalent secret key streams can be (partially) recovered under plaintext attacks. Taking the above analyses as guidelines for design, we propose a new chaotic image cipher with SPN structure using data-dependent cryptographical operations. Extensive analyses and numerical simulations are carried out to validate the superiority of this design.
In the second part of the thesis, some selective encryption techniques based on a joint sampling and source-cryptographical coding scheme, called compressed sensing (CS), are investigated. The theoretical secrecy upper bound that can be achieved by the basic CS paradigm is first reviewed and the security flaws of some recent proposed CS-based product ciphers are then studied. To bypass the secrecy upper bound and fix the security flaws of the existing designs, a bi-level protected CS (BLP-CS) model is suggested by embedding a transform-domain encryption to the CS paradigm. It is theoretically shown that BLP-CS is computationally secure against plaintext attacks, while existing schemes are not. Lastly, analogous to selective encryptions of multimedia data that are Shannon-Nyquist sampled, selective encryption of compressively sensed data are examined by studying the interplay of traditional ciphers and CS.