Jiajin Le

Providing privacy preserving in cloud computing

People can only enjoy the full benefits of Cloud computing if we can address the very real privacy and security concerns that come along with storing sensitive personal information in databases and software scattered around the Internet. There are many service provider in the internet, we can call each service as a cloud, each cloud service will exchange data with other cloud, so when the data is exchanged between the clouds, there exist the problem of disclosure of privacy. So the privacy disclosure problem about individual or company is inevitably exposed when releasing or sharing data in the cloud service. Privacy is an important issue for cloud computing, both in terms of legal compliance and user trust, and needs to be considered at every phase of design. Our paper provides some privacy preserving technologies used in cloud computing services.

A Survey on Privacy Preserving Data Mining

Privacy preserving becomes an important issue in the development progress of data mining techniques. Privacy preserving data mining has become increasingly popular because it allows sharing of privacy-sensitive data for analysis purposes. So people have become increasingly unwilling to share their data, frequently resulting in individuals either refusing to share their data or providing incorrect data. In turn, such problems in data collection can affect the success of data mining, which relies on sufficient amounts of accurate data in order to produce meaningful results. In recent years, the wide availability of personal data has made the problem of privacy preserving data mining an important one. A number of methods have recently been proposed for privacy preserving data mining of multidimensional data records. This paper intends to reiterate several privacy preserving data mining technologies clearly and then proceeds to analyze the merits and shortcomings of these technologies.

A Survey on Privacy Preserving Approaches in Data Publishing

Privacy preserving in data publishing has become one of the most important research topics in data security field and it has become a serious concern in publication of personal data in recent years. How to efficiently protect individual privacy in data publishing is especially critical. Thus, various proposals have been designed for privacy preserving in data publishing. In this paper, we summarize privacy preserving approaches in data publishing and survey current existing techniques, and analyze the advantage and disadvantage of these approaches. We divide these proposals into two categories, one is to achieve the purpose of privacy preserving based on k-anonymity model, and the other is to utilize the methods of probability or statistics to protect data privacy in the case of the statistical properties of the final data and classification properties are unchanged. For example, clustering, randomization approaches. Finally, we discuss the future directions of privacy preserving in data publishing.

Effective automatic image annotation via integrated discriminative and generative models

In this paper, we present a novel image annotation method that leverages on the advantages of both generative and discriminative models. To label an image, we first identify a visual neighborhood in the training image set based on generative approach. Then, the neighborhood is refined by an optimal discriminative hyperplane tree classifier based on concept feature. The tree classifier is built according to a local topic hierarchy, which is adaptively constructed by exploiting the semantic contextual correlations of the corresponding visual neighborhood. Experiments conducted on the ECCV2002 and TRECVID 2005 benchmarks demonstrate the effectiveness and efficiency of the proposed method.

Biologically-Inspired Aging Face Recognition Using C1 and Shape Features

To deal with the variations caused by age, an aging face recognition method Based on HMAX model, which motivated by a quantitative model of visual cortex, was proposed to achieve temporal invariance. First, each face image was normalized to a standard size. Second, the C1-S features, which preserve facial texture and shape information, were defined by facial key points and HMAX model to represent the face image with the high dimensional features. Then C1-S features are projected to a low dimensional subspace by PCA. Finally, the nearest neighbor rule with Mahalanobis distance was used to aging face recognition from rank 1 to rank 6. Experiments on the FG-NET database show that our proposed C1-S features are good at tolerating local position, scale and aging variations and improve the accuracy of aging face recognition.

Improving CLOPE’s Profit Value and Stability with an Optimized Agglomerative Approach

CLOPE (Clustering with sLOPE) is a simple and fast histogram-based clustering algorithm for categorical data. However, given the same data set with the same input parameter, the clustering results by this algorithm would possibly be different if the transactions are input in a different sequence. In this paper, a hierarchical clustering framework is proposed as an extension of CLOPE to generate stable and satisfactory clustering results based on an optimized agglomerative merge process. The new clustering profit is defined as the merge criteria and the cluster graph structure is proposed to optimize the merge iteration process. The experiments conducted on two datasets both demonstrate that the agglomerative approach achieves stable clustering results with a better profit value, but costs much more time due to the worse complexity.

Classifying children's and adults' faces by bio-inspired features

Children are usually treated differently from adults in many computer vision applications. To classify children from adults by face images in a natural and non-intrusive way, a method using improved bio-inspired features (C1-S) is presented in this paper. To reduce the negative influence of individual differences, active shape model (ASM) is used to extract 58 landmarks for face normalization. Motivated by quantitative model of visual cortex, we proposed C1-S features to represent each face. The features output from C1 units consider not only the points defined by grid size but also the points defined by ASM fitting results. By adding shape features, C1-S features have better performance in SVM classification. Experiment results show that our method provides good classification accuracy and can be used for home video surveillance and parental control.

(α, β, k)-anonymity: An effective privacy preserving model for databases

Publishing the data with multiple sensitive attributes brings us greater challenge than publishing the data with single sensitive attribute in the area of privacy preserving. In this paper, we propose a novel privacy preserving model based on k-anonymity called (α, β, k)-anonymity for databases. (α, β, k)-anonymity can be used to protect data with multiple sensitive attributes in data publishing. Then, we set a hierarchy sensitive attribute rule to achieve (α, β, k)-anonymity model and develop the corresponding algorithm to anonymize the microdata by using generalization and hierarchy. We verify (α, β, k)-anonymity approach can effectively protect privacy information of individual and resist background knowledge attack in publishing the data with multiple sensitive attributes by specific example.

A Dynamic Logging Approach on Flash-Based DBMS

Flash memory, especially NAND flash memory, is being rapidly deployed as data storage for mobile devices such as mobile phones, digital cameras and PDApsilas. With its capacity increasing and price dropping, flash memory has been installed in many portable computers instead of magnetic disk for its light weight, small size, physical stability and low power consumption. We may expect that flash memory will finally take the place of magnetic disk in the next few years. Therefore, it is possible for us to consider running a full database on the flash computing platforms. However, disk-based database technology can not be used on flash memory directly because of different characteristics between flash memory and magnetic disk. Some researches are being done currently on flash-based database technology. In this paper, we propose a new approach called dynamic logging. This new approach makes use of the characteristics of flash memory effectively to acquire more benefits from using traditional database technology on flash memory.

Learning Bayesian Network Structure from Large-Scale Datasets

Bayesian network is one of the most classical and effective models in big data graph algorithms. Aiming at the problem of learning Bayesian network structure from large-scale datasets, a novel algorithm with the combination of Information theory, Tabu search and Akaike Information Criterion (AIC) called ITA is proposed. Firstly, a dimensionreduction algorithm based on information theory is used to filter non-target variables. The variables closely related to the target are picked as the vertexes in Bayesian network. Then choosing AIC as the scoring method and Tabu Search as the heuristic algorithm, a new learning algorithm is adopted to build the global optimal structure. Experimental results demonstrate that ITA algorithm can obtain core causal relationships from largescale datasets in certain area accurately and construct clean and straightforward Bayesian network structure at a lower time cost. Therefore, ITA is an effective and efficient big data graph algorithm for learning Bayesian network structure from largescale datasets.