Biblio

Nowadays, biometric-based authentication systems are widely used. This fact has led to increased attacks on biometric data of users. Therefore, biometric template protection is sure to keep the attention of researchers for the security of the authentication systems. Many previous works proposed the biometric template protection schemes by transforming the original biometric data into a secure domain, or establishing a cryptographic key with the use of biometric data. The main purpose was that fulfill the all three requirements: cancelability, security, and performance as many as possible. In this paper, using random projection merged with fuzzy commitment, we will introduce a hybrid scheme of biometric template protection. We try to limit their own drawbacks and take full advantages of these techniques at the same time. In addition, an analysis of non-invertibility property will be exercised with regards to the use of random projection aiming at enhancing the security of the system while preserving the discriminability of the original biometric template.

Matrix sketching is aimed at finding compact representations of a matrix while simultaneously preserving most of its properties, which is a fundamental building block in modern scientific computing. Randomized algorithms represent state-of-the-art and have attracted huge interest from the fields of machine learning, data mining, and theoretic computer science. However, it still requires the use of the entire input matrix in producing desired factorizations, which can be a major computational and memory bottleneck in truly large problems. In this paper, we uncover an interesting theoretic connection between matrix low-rank decomposition and lossy signal compression, based on which a cascaded compression sampling framework is devised to approximate an m-by-n matrix in only O(m+n) time and space. Indeed, the proposed method accesses only a small number of matrix rows and columns, which significantly improves the memory footprint. Meanwhile, by sequentially teaming two rounds of approximation procedures and upgrading the sampling strategy from a uniform probability to more sophisticated, encoding-orientated sampling, significant algorithmic boosting is achieved to uncover more granular structures in the data. Empirical results on a wide spectrum of real-world, large-scale matrices show that by taking only linear time and space, the accuracy of our method rivals those state-of-the-art randomized algorithms consuming a quadratic, O(mn), amount of resources.

In order to achieve more secure and privacy-preserving, a new method of cancelable palmprint template generation scheme using noise data is proposed. Firstly, the random projection is used to reduce the dimension of the palmprint image and the reduced dimension image is normalized. Secondly, a chaotic matrix is produced and it is also normalized. Then the cancelable palmprint feature is generated by comparing the normalized chaotic matrix with reduced dimension image after normalization. Finally, in order to enhance the privacy protection, and then the noise data with independent and identically distributed is added, as the final palmprint features. In this article, the algorithm of adding noise data is analyzed theoretically. Experimental results on the Hong Kong PolyU Palmprint Database verify that random projection and noise are generated in an uncomplicated way, the computational complexity is low. The theoretical analysis of nosie data is consistent with the experimental results. According to the system requirement, on the basis of guaranteeing accuracy, adding a certain amount of noise will contribute to security and privacy protection.

A robust appearance model is usually required in visual tracking, which can handle pose variation, illumination variation, occlusion and many other interferences occurring in video. So far, a number of tracking algorithms make use of image samples in previous frames to update appearance models. There are many limitations of that approach: 1) At the beginning of tracking, there exists no sufficient amount of data for online update because these adaptive models are data-dependent and 2) in many challenging situations, robustly updating the appearance models is difficult, which often results in drift problems. In this paper, we proposed a tracking algorithm based on compressive sensing theory and particle filter framework. Features are extracted by random projection with data-independent basis. Particle filter is employed to make a more accurate estimation of the target location and make much of the updated classifier. The robustness and the effectiveness of our tracker have been demonstrated in several experiments.

Nearest neighbor search is a fundamental problem in various research fields like machine learning, data mining and pattern recognition. Recently, hashing-based approaches, for example, locality sensitive hashing (LSH), are proved to be effective for scalable high dimensional nearest neighbor search. Many hashing algorithms found their theoretic root in random projection. Since these algorithms generate the hash tables (projections) randomly, a large number of hash tables (i.e., long codewords) are required in order to achieve both high precision and recall. To address this limitation, we propose a novel hashing algorithm called density sensitive hashing (DSH) in this paper. DSH can be regarded as an extension of LSH. By exploring the geometric structure of the data, DSH avoids the purely random projections selection and uses those projective functions which best agree with the distribution of the data. Extensive experimental results on real-world data sets have shown that the proposed method achieves better performance compared to the state-of-the-art hashing approaches.

Vector space models (VSMs) are mathematically well-defined frameworks that have been widely used in text processing. In these models, high-dimensional, often sparse vectors represent text units. In an application, the similarity of vectors -- and hence the text units that they represent -- is computed by a distance formula. The high dimensionality of vectors, however, is a barrier to the performance of methods that employ VSMs. Consequently, a dimensionality reduction technique is employed to alleviate this problem. This paper introduces a new method, called Random Manhattan Indexing (RMI), for the construction of L1 normed VSMs at reduced dimensionality. RMI combines the construction of a VSM and dimension reduction into an incremental, and thus scalable, procedure. In order to attain its goal, RMI employs the sparse Cauchy random projections.