Detection of Denial of Service Attacks Using J48 Algorithm Compared with Naive Bayes Algorithm to Improve Positive Detection Rate

Abstract

Aim: The goal of this study is to compare accuracy to evaluate the efficiency of two machine learning methods for detecting Denial of Service (DOS) assaults.Materials and Methods: To diagnose DoS assaults and obtain a prediction rate to compare algorithms, several Novel J48 algorithms were used as group 1 and Naive bayes algorithms as group 2. The algorithm should be capable of detecting the specific type of DoS attack. For each of the groups studied, a sample size of N=20 was evaluated for implementation. SPSS was used to calculate the sample size. The pre-test analysis was maintained at 80%. G-power is used to calculate sample size. Results: Based on statistical analysis, the significance value for calculating  accuracy  was found to be 0.046.  The significant values for calculating Flow Duration and Idle Std were found to be 0.945 and 0.266, respectively, based on statistical analysis. With a mean Flow Duration percentage of 91.14 percent, the Novel J48 Method is somewhat more accurate than the Naive Bayes algorithm, which has a mean Flow Duration percentage of 86.48 percent. Conclusion: The Novel J48 method provides a slightly better prediction rate value than the Naive Bayes  technique when it comes to detecting DoS assaults.