Anomaly Based Intrusion Detection in Mixed Attribute Dataset Using Data Mining Methods
B. A. Manjunatha, Prasanta Gogoi · Journal of Artificial Intelligence · 2015
Background: In network security system anomaly detection is extremely important part in mixed attribute dataset.Detection of unexpected behavior in the network, by using outdated methods anomaly detection becomes inapt because data naturally occurs as mixture of numerical and categorical attributes.Methodology: The proposed algorithm, Minimum Threshold Support Count (MTSC) and modified Canberra method is used to detect mainly anomalies in categorical and numerical attributes (mixed attributes) that deals with sparse high-dimensionality of currently available dataset.Enhanced adaptive boosting classifier is very sensitive to anomalies and infrequent data.The accuracy and performance of the proposed method is comprehensively improved by using enhanced adaptive boosting classifier.Results: Results show that the classification True Positive Rate (TPR), precision, recall, F-measure and ROC area are more and false positive rate is less for the proposed method when compared to existing method.Conclusion: Proposed method gives effective classification accuracy and less computation time.