It is an important task for an Intrusion Detection System (IDS) to identify abnormal user behaviors accurately and quickly. In order to solve the problems of high dimensionality and large sample size of intrusion detection data, a related family attribute reduction method based on variable precision covering rough set was proposed, and was applied to the intrusion detection data. Firstly, the variable precision related families with condition attributes were generated based on the covering decision table. Then, a heuristic algorithm was used to obtain the attribute reduction of the decision table based on all the variable precision related families with condition attributes. Finally, the intrusion detection data was detected by combining with the classifier on the above basis. Experimental results show that, the proposed method has the low time complexity of calculating attribute reduction, and on large sample datasets, the running time of attribute reduction algorithm named Neighborhood Fuzzy Rough Sets (NFRS) based on fuzzy rough set dependency is 96 times of that of the proposed method. On the NSL-KDD dataset, the proposed method can identify key attributes quickly, eliminate invalid information, and has the overall accuracy reached 90.53% and the accuracy of Normal reached 97%.

Deep learning has achieved good results in human action recognition, but it still needs to make full use of video human appearance information and motion information. To recognize human actions by using spatial information and temporal information in video, a video human action recognition model based on spatio-temporal two-stream was proposed. Two convolutional neural networks were used to extract spatial and temporal features of video sequences respectively in the proposed model, and then the two neural networks were merged to extract the middle spatio-temporal features, finally the video human action recognition was completed by inputting the extracted features into a 3D convolutional neural network. The video human action recognition experiments were carried out on the data set UCF101 and HMDB51. Experimental results show that the proposed 3D convolutional neural network model based on the spatio-temporal two-stream can effectively recognize the video human actions.

Aiming at the "uniform effect" of the traditional K-means algorithm, a new probability model-based algorithm was proposed for non-uniform data clustering. Firstly, a Gaussian mixture distribution model was proposed to describe the clusters hidden within non-uniform data, allowing the datasets to contain clusters with different densities and sizes at the same time. Secondly, the objective optimization function for non-uniform data clustering was deduced based on the model, and an EM (Expectation Maximization)-type clustering algorithm defined to optimize the objective function. Theoretical analysis shows that the new algorithm is able to perform soft subspace clustering on non-uniform data. Finally, experimental results on synthetic datasets and real datasets demostrate that the accuracy of the proposed algorithm is increased by 5% to 50% compared with the existing K-means-type algorithms and under-sampling algorithms.

Aiming at the problem that the current K-means-type soft-subspace algorithms cannot effectively cluster imbalanced data due to uniform effect, a new partition-based algorithm was proposed for soft subspace clustering on imbalanced data. First, a bi-weighting method was proposed, where each attribute was assigned a feature-weight and each cluster was assigned a cluster-weight to measure its importance for clustering. Second, in order to make a trade-off between attributes with different types or those categorical attributes having various numbers of categories, a new distance measurement was then proposed for mixed-type data. Third, an objective function was defined for the subspace clustering algorithm on imbalanced data based on the bi-weighting method, and the expressions for optimizing both the cluster-weights and feature-weights were derived. A series of experiments were conducted on some real-world data sets and the results demonstrated that the bi-weighting method used in the new algorithm can learn more accurate soft-subspace for the clusters hidden in the imbalanced data. Compared with the existing K-means-type soft-subspace clustering algorithms, the proposed algorithm yields higher clustering accuracy on imbalanced data, achieving about 50% improvements on the bioinformatic data used in the experiments.

As standard Backtracking Search Optimization Algorithm (BSA) has the shortcoming of slow convergence, a new mutation scale factor based on Maxwell-Boltzmann distribution and a crossover strategy with greedy property were introduced to improve it. Maxwell-Boltzmann distribution was used to generate mutation scale factor, which could enhance search efficiency and convergence speed. Mutation population learning from outstanding individuals was adopted in less exchange-dimensional crossover strategy to add greedy property to crossover as well as fully ensure population diversity, which managed to avoid the problem that most existed algorithms easily trap into local minima when added greedy property. The simulation experiments were conducted on fifteen Benchmark functions. The results show that the improved algorithm has faster convergence speed and higher convergence precision, even in the high-dimensional multimodal functions, the improved algorithm's search results are nearly 14 orders of magnitude higher than those of original BSA after the same iterations, and its convergence precision can reach 10-10 or less.

Based on the principle and characteristics of the Differential Power Analysis (DPA) attack, the kernel function was used to estimate the probability distribution density of the leakage of power consumption in the password chip work process. By calculating the mutual information between the attack model and the power leakage, when the guessed key was correct, this paper quantified the risk value of the password chip in the face of DPA attacks. The experiments show that the risk quantification method can be a good estimate of the correlation degree between the attack model and power leakage when the guessed key is correct and then provides important indicators to complete password chip risk evaluation.

In order to solve the problem that traditional classification is not very satisfactory due to fewer text features in short text, an algorithm using concept description was presented. At first, a global semantic concept word list was built. Then the test set and training set were conceptualized by the global semantic concept word list to combine the test short texts by the same description of concept in the training set, and at the same time, training long texts were combined by the training short texts in the training set. At last, the long text was classified by traditional classification algorithm. The experiments show that the proposed method could mine implicit semantic information in short text efficiently while expanding short text on semantics adequately, and improving the accuracy of short text classification.

The recognition results of the same target by different sensors are often contradictory in wireless sensor networks. The use of data fusion based on DempsteShafer (D-S) evidence theory could solve this problem. However, when using D-S evidence combination formula to compute，with the increase of the target identity, the computation will be growing rapidly. The processing ability of sensor nodes is limited and the data of decision in sensor networks are redundant, thus，a way was proposed to reduce the number of target identity by preprocessing and to reduce the computation; and it could rule out the data with errors through greater consistency test; therefore，it makes decision results more accurate.

Current intrusion detection systems lack the ability to generalize from previously observed attacks and to detect even slight variations of known attacks. An approach employing LS and neural networks was described to provide the ability to generalize from previously observed behavior and to recognize future unseen behavior. The method was represented to use feedback neural networks in anomaly detection to structure the characteristic pattern of the short sequences of system calls. Meanwhile, the algorithm and design of the neural network were given. Experiment shows that the neural network is especially better to deal with events and variance of intrusions and improves the detection rate without increasing the false positives.