Detecting masquerades using a combination of Naïve Bayes and weighted RBF approach

Abstract

Masquerade detection by automated means is gaining widespread interest due to the serious impact of masquerades on computer system or network. Several techniques have been introduced in an effort to minimize up to some extent the risk associated with masquerade attack. In this respect, we have developed a novel technique which comprises of Naïve Bayes approach and weighted radial basis function similarity approach. The proposed scheme exhibits very promising results in comparison with many earlier techniques while experimenting on SEA dataset in detecting masquerades.

Adaptive Naive Bayes method for masquerade detection

“…Lane 18 proposes a model of intrusion detection by adopting a Partially Ordered Markov Decision Process (POMDP) for user profiling, where the classifier is trained using partly labeled data and partly unlabeled data. Recently, a combined weighted Radial Basis Function (RBF)‐Naive Bayes algorithm is proposed by Sharma and Paliwal 19. This algorithm uses weighted RBF similarity measure with Naive Bayes technique for classification.…”

“…Table I lists the best performance of each of the algorithms, in terms of the best combination of hit rate and FPR, as reported by respective papers. Some of the authors, like Sharma and Paliwal 19, present the result of relative performance of their algorithm in comparison to some of the existing algorithms by fixing one parameter and comparing on the basis of the other. Absolute performance measure of their algorithm is not reported.…”

The Incremental Probabilistic Neural Network

“…and W is the weight matrix of output layer, 2 B is the bias vector of output layer, 1 A is the output vector of the hidden layer, I is a 1×1 vector which the element is 1,…”

“…Construct the output layer by hidden layer output. Calculate the weight vector of output layer through 3 1 ) ( T WA MAX = , and 3 T is the target vector of the PNN, W is the weight matrix of the output layer, 1 A is the output vector of the hidden layer.…”

“…For detecting masquerades attacks, Alok Sharma et al [1] proposed a new technique combining Naïve Bayes approach and weighted Radial Basis Function (RBF) similarity approach. Dimitris Gavrilis et al [2] designed a RBFNN detector for DDoS (Distributed Denial-Of-Service) attacks.…”