Improving the Accuracy of Intrusion Detection System in the Detection of DoS using Naive Bayes with Lasso Feature Elimination and Comparing with Naive Bayes without Feature Elimination in Wireless Adhoc Network

Abstract

The aim of this research is to create an InnovativeNaive Bayes with Lasso Feature Elimination Intrusion Detection System (IDS) that uses Naive Bayes without feature elimination (Group 1) and compare its performance to that with Lasso feature elimination (Group 2). NSL-KDD Dataset was used to design the data set and collect an IDS. A total of 38 samples were obtained from each of the 19 groups. The data was analyzed using the SPSS application for statistical analysis. Both groups were subjected to an independent sample T test, which yielded a significance of 0.595 for accuracy. Here p > 0.05. For Group 1, the mean accuracy of Naive Bayes without feature elimination is 0.7432, and for Group 2, the mean accuracy of Lasso feature elimination is 0.6005. Conclusion: The accuracy of the Naive Bayes with Lasso feature elimination is similar to that of the Naive Bayes without feature elimination, but here significance is existing.