Probability Distributions.- Linear Models for Regression.- Linear Models for Classification.- Neural Networks.- Kernel Methods.- Sparse Kernel Machines.- Graphical Models.- Mixture Models and EM.- Approximate Inference.- Sampling Methods.- Continuous Latent Variables.- Sequential Data.- Combining Models.

Pattern Recognition and Machine Learning

Compressive Sensing (CS) is a new sensing modality, which compresses the signal being acquired at the time of sensing. Signals can have sparse or compressible representation either in original domain or in some transform domain. Relying on the sparsity of the signals, CS allows us to sample the signal at a rate much below the Nyquist sampling rate. Also, the varied reconstruction algorithms of CS can faithfully reconstruct the original signal back from fewer compressive measurements. This fact has stimulated research interest toward the use of CS in several fields, such as magnetic resonance imaging, high-speed video acquisition, and ultrawideband communication. This paper reviews the basic theoretical concepts underlying CS. To bridge the gap between theory and practicality of CS, different CS acquisition strategies and reconstruction approaches are elaborated systematically in this paper. The major application areas where CS is currently being used are reviewed here. This paper also highlights some of the challenges and research directions in this field.

A Systematic Review of Compressive Sensing: Concepts, Implementations and Applications

The applications of compressive sensing (CS) in the field of information security have captured a great deal of researchers’ attention in the past decade. To supply guidance for researchers from a comprehensive perspective, this paper, for the first time, reviews CS in information security field from two aspects: theoretical security and application security. Moreover, the CS applied in image cipher is one of the most widespread applications, as its characteristics of dimensional reduction and random projection can be utilized and integrated into image cryptosystems, which can achieve simultaneous compression and encryption of an image or multiple images. With respect to this application, the basic framework designs and the corresponding analyses are investigated. Specifically, the investigation proceeds from three aspects, namely, image ciphers based on chaos and CS, image ciphers based on optics and CS, and image ciphers based on chaos, optics, and CS. A total of six frameworks are put forward. Meanwhile, their analyses in terms of security, advantages, disadvantages, and so on are presented. At last, we attempt to indicate some other possible application research topics in future.

https://dro.deakin.edu.au/eserv/DU:30106242/zhang-reviewofcompressive-2016.pdf

A Review of Compressive Sensing in Information Security Field

A visually secure image encryption scheme based on parallel compressive sensing

An image coding scheme using parallel compressive sensing for simultaneous compression-encryption applications

Introducing the counter mode of operation to Compressed Sensing based encryption

Compressed Sensing may offer confidentiality with a good level of security that comes at limited extra costs. By this means, encryption can be added right into the sampling process. Previous proposals of Compressive Sensing encryption schemes are based on the randomness introduced by the sensing matrix. Once the matrix is fixed, the encryption becomes deterministic and therefore it can hardly be secure when encrypting more than a single signal. The work presented in this paper is built on the theoretical results on the secrecy of one-time Compressive Sensing encryption, and it proposes practical realizations, which extend this kind of encryption so that multiple signals may be encrypted under a single key. The confidentiality of the signal is thereby solely given by the inherent secrecy of the compressed measurements. In contrast to recent suggestions, the modes of operation defined in this paper are able to encrypt signals with different energy without exposing the signals energy to an eavesdropper. A general design for Compressive Sensing encryption modes is presented along with two realizations: one designed for parallel processing and another one that is self-synchronizing. The secrecy of the proposed modes is reduced to the secrecy of known and trusted cryptographic primitives. Compressive Sensing encryption modes are useful in a wide range of practical applications by providing a joint end-to-end encryption and compression that starts at the sensor level.

Compressive Sensing encryption modes and their security

Feature extraction is the most critical part of biometric authentication systems. The majority of biometric systems proposed in the last years are using alignment to ensure robust authentication in the presence of affine transformations like rotation and translation. Nevertheless, alignment is time consuming, and misalignment leads to the lack of accuracy. Using template-protection, there is a need for additional information to perform explicit alignment. It is therefore not clear whether this information could be used to attack the protected biometric template. This Chapter presents a comparative view on alignment-free features for biometric authentication from the perspective of pattern recognition and digital image processing as well as biometrics. The basics of these disciplines are aggregated and different proposed techniques are described, assessed and compared. Finally, an evaluation strategy from the field of digital image processing is applied to biometrics in order to assess robustness and invariance of feature extraction in biometrics.

Robin Fay

Papers

Introducing the counter mode of operation to Compressed Sensing based encryption

Compressive Sensing encryption modes and their security

Robustness of Biometrics by Image Processing Technology