Vladimir Golovko

Dimensionality Reduction and Attack Recognition using Neural Network Approaches

Most current Intrusion Detection Systems (IDS) examine all data features to detect intrusion. Also existing intrusion detection approaches have some limitations, namely impossibility to process a large number of audit data for realtime operation, low detection and recognition accuracy. To overcome these limitations, we apply modular neural network models to detect and recognize attacks in computer networks. They are based on the combination of principal component analysis (PCA) neural networks and multilayer perceptrons (MLP). PCA networks are employed for important data extraction and to reduce high dimensional data vectors. We present two PCA neural networks for feature extraction: linear PCA (LPCA) and nonlinear PCA (NPCA). MLP is employed to detect and recognize attacks using feature-extracted data instead of original data. The proposed approaches are tested with the help of KDD-99 dataset. The experimental results demonstrate that the designed models are promising in terms of accuracy and computational time for real world intrusion detection.

Neural Network and Artificial Immune Systems for Malware and Network Intrusion Detection

Neural network techniques and artificial immune systems (AIS) have been successfully applied to many problems in the area of anomaly activity detection and recognition. The existing solutions use mostly static approaches, which are based on collection viruses or intrusion signatures. Therefore the major problem of traditional techniques is detection and recognition of new viruses or attacks. This chapter discusses the use of neural networks and artificial immune systems for intrusion and virus detection. We studied the performance of different intelligent techniques, namely integration of neural networks and AIS for virus and intrusion detection as well as combination of various kinds of neural networks in modular neural system for intrusion detection. This approach has good potential to recognize novel viruses and attacks.

Neural Network Ensembles for Intrusion Detection

The major problem of existing models is recognition of new attacks, low accuracy, detection time and system adaptability. In this paper the method of recognition of attack class on the basis of the analysis of the network traffic is described. Our first approach is based on combination principal component analysis (PCA) neural networks and multilayer perceptrons (MLP). The second approach performs recognition of a class of attack by means of the cumulative classifier with nonlinear recirculation neural networks (RNN) as private detectors. The proposed approaches are tested using KDD-99 dataset. The experimental results demonstrate that the designed models are promising in terms of accuracy and computational time for real world intrusion detection.

Neural network approach to real-time network intrusion detection and recognition

Modern intrusion detection systems process large amounts of data. Most systems use signature- and rule-based approaches to find attack traces. The main disadvantage of such technologies is the need of continuous updating of signature database to let the system detect newest attacks. We present recirculation neural network based approach which lets to detect previously unseen attack types in real-time mode and to further correct recognition of this types. The experiments held on both KDD data and real network traffic data prove that this approach can be used in host-based anomaly and misuse detectors.

Intrusion Recognition Using Neural Networks

Intrusion detection techniques are of great importance for computer network protecting because of increasing the number of remote attack using TCP/IP protocols. There exist a number of intrusion detection systems, which are based on different approaches for anomalous behavior detection. This paper focuses on applying neural networks for attack recognition. It is based on multilayer perceptron. The 1999 KDD Cup data set is used for training and testing neural networks. The results of experiments are discussed in the paper.

Development of neural network immune detectors for computer attacks recognition and classification

There is developed a combined method that is based on the integration of neural network detectors in an artificial immune system. This allowed them to adapt to new attacks with the help of cloning and mutation operations.

Intelligent system for detection of networking intrusion

A modified neural detector and a method of principal components analysis were considered to improve a quality of network attacks detection.

A Learning Technique for Deep Belief Neural Networks

Deep belief neural network represents many-layered perceptron and permits to overcome some limitations of conventional multilayer perceptron due to deep architecture. The supervised training algorithm is not effective for deep belief neural network and therefore in many studies was proposed new learning procedure for deep neural networks. It consists of two stages. The first one is unsupervised learning using layer by layer approach, which is intended for initialization of parameters (pre-training of deep belief neural network). The second is supervised training in order to provide fine tuning of whole neural network. In this work we propose the training approach for restricted Boltzmann machine, which is based on minimization of reconstruction square error. The main contribution of this paper is new interpretation of training rules for restricted Boltzmann machine. It is shown that traditional approach for restricted Boltzmann machine training is particular case of proposed technique. We demonstrate the efficiency of proposed approach using deep nonlinear auto-encoder.

Application of Neural Networks to the Electroencephalogram Analysis for Epilepsy Detection

Many techniques were used in order to detect and to predict epileptic seizures on the basis of electroencephalograms. One of the approaches for the prediction of the epileptic seizures is the use the chaos theory, namely determination largest Lyapunovs exponent or correlation dimension of the scalp EEG signals. This paper presents the neural network technique for epilepsy detection. It is based on computing of the largest Lyapunovs exponent. The results of experiments are discussed.

A new technique for restricted Boltzmann machine learning

Over the last decade, deep belief neural networks have been a hot topic in machine learning. Such networks can perform a deep hierarchical representation of input data. The first layer can extract low-level features, the second layer can extract high-level features and so on. In general, deep belief neural network represents many-layered perceptron and permits to overcome some limitations of conventional multilayer perceptron due to deep architecture. In this work we propose a new training technique called Reconstruction Error-Based Approach (REBA) for deep belief neural network based on restricted Boltzmann machine. In contrast to classical Hintons training approach, which is based on a linear training rule, the proposed technique is based on a nonlinear learning rule. We demonstrate the performance of REBA technique for the MNIST dataset visualization. The main contribution of this paper is a novel view on the training of a restricted Boltzmann machine.