Search

High precision and low false alarm rate are the two most important characteristics of a good Intrusion Detection System (IDS). In this work, we propose to construct a host-based IDS for detecting flooding-based Denial of Service (DoS) attacks by minimizing the generalization error bound of the IDS to reduce its false alarm rate and increase its precision. Radial basis function neural network (RBFNN) will be applied in the IDS. The generalization error bound is formulated based on the stochastic sensitivity measure of RBFNN. Experimental results using artificial datasets support our claims. [ABSTRACT FROM AUTHOR]

Detecting multiple network attacks is essential to intrusion detection, network security defense and network traffic management. This paper presents a covariance matrix based detection approach to detecting multiple known and unknown network anomalies. It utilizes the difference of covariance matrices among observed samples in the detection. A threshold matrix is employed in the detection where each entry of the matrix evaluates the covariance changes of the corresponding features. As case studies, extensive experiments are conducted to detect multiple DoS attacks - the prevalent Internet anomalies. The experimental results indicate that the proposed approach achieves high detection rates in detecting multiple known and unknown anomalies. [ABSTRACT FROM AUTHOR]

The network security problem has become a critical issue and many approaches have been proposed to tackle the information security problems, especially the Denial of Service (DoS) attacks. Multiple Classifier System (MCS) is one of the approaches that have been adopted in the detection of DoS attacks recently. Fusion strategy is crucial and has great impact on the classification performance of an MCS. However the selection of the fusion strategy for an MCS in DoS problem varies widely. In this paper, we focus on the comparative study on adopting different fusion strategies for an MCS in DoS problem. [ABSTRACT FROM AUTHOR]

In this paper the definition of local reduction is proposed to describe the minimal description of a definable set by attributes of the given information system. The local reduction can present more optimal description for single decision class than the existing relative reductions. It is proven that the core of reduction or relative reduction can be expressed as the union of the cores of local reductions. The discernibility matrix of reduction and relative reduction can be obtained by composing discernibility matrixes of local reduction. [ABSTRACT FROM AUTHOR]

Many methods based on rough sets to deal with incomplete information system have been proposed in recent years. However, they are only suitable for the nominal datasets. So far only a few methods based on rough sets to deal with incomplete information system with continuous-valued attributes have been proposed. In this paper we propose one generalized model of rough sets to reduce continuous-valued attributes and learn some rules in an incomplete information system. The definition of a relative discernible measure is firstly proposed, which is the underlying concept to redefine the concepts of knowledge reduction such as the reduct and core. We extend a number of underlying concepts of knowledge reduction (such as the reduct and core), and finally propose a heuristic algorithm to generate fuzzy reduct from initial data. The main contribution of this paper is that the underlying relationship between the reduct and core of rough sets is proved to be still correct after our extension. The advantage of the proposed method is that instead of preprocessing continuous data by discretization or fuzzification, we can reduce an incomplete information system with continuous-valued attributes directly based on the generalized model of rough sets, Finally a numerical example is given to show the feasibility of our proposed method. [ABSTRACT FROM AUTHOR]

In this paper we develop a fuzzy neural network (FNN) with a new BP learning algorithm using some smooth function, which is used to tune the local and global weights of fuzzy production rules (FPRs) so as to enhance the representation power of FPRs. The aim of including local and global weights in FPRs and the tuning of these weights is to improve the learning and testing accuracy without increasing the number of rules. By experimenting with some existing benchmark examples (Iris data, Wine data, Pima data and Glass data) the proposed method is found to have high accuracy in classifying unseen samples without increasing the number of the extracted FPRs, and furthermore, the time required to consult with domain experts for gaining a rule is reduced. The synergy between WFPRs and an FNN offers a new insight into the construction of better fuzzy intelligent systems in the future. [ABSTRACT FROM AUTHOR]