Accurate Deauthentication Attack Detection using Linear Discriminant Analysis in Comparison with Multilayer Perceptron

Abstract

Aim: This study aims to compare the accuracy of Linear Discriminant Analysis to Multilayer Perceptron in detecting deauthentication attack. Materials and Methods: 61,000 samples were taken for this analysis with two data sets in the ratio of 80% and 20% samples each. 80% data set is used for training the model and 20% data is used for testing. By the workflow of this research, the data set has been imported, LDA and MLP code has been implemented with the help of jupyter notebooks in the google colab platform. The sample size is calculated from the values obtained from the previous studies with help of the online statistical analysis tool with confidence level of 95% and the margin error of 1%. Results: The accuracy of LDA was 82.25% P<0.05, whereas the accuracy of the MLP was 86.48% P<0.05. Conclusion: For the given data set MLP (Multilayer Perceptron) performs significantly better than LDA (Linear Discriminant Analysis) in detecting the deauthentication attack