A Comparative Study of LAD, CNN and DNN for Detecting Intrusions

Sneha Chauhan, Loreen Mahmoud, Sugata Gangopadhyay, Aditi Kar Gangopadhyay

Abstract: In recent years, with the growth of the Internet and network devices, a significant amount of information has been exposed to the attackers and intruders. Due to vulnerabilities in the system, the adversaries plan new ways of network intrusions. Many Intrusion Detection Systems (IDSs) are developed to protect the networks from malicious attacks and ensure reliability and availability within the organizations. IDSs built using various machine learning and data mining techniques are effective in detecting attacks. However, their performance decreases with an increase in the size of data. In this paper, we focus on developing an IDS model using Logical Analysis of Data (LAD). It is a supervised learning technique where patterns are generated using partially defined Boolean functions (pdBf), which can detect attacks based on certain features of the data. We compare the performance of LAD model with Deep Neural Network (DNN) and Convolutional Neural Network (CNN) IDS models. UNSW-NB15 and CSE-CIC-IDS2018 datasets are used for training and testing our proposed model. The results show that the performance of LAD model is competitive to CNN, DNN and other existing IDS models based on accuracy, precision, recall and F1 score.