Highly accurate anomaly based intrusion detection through integration of the local outlier factor and convolutional neural network

Abstract

Intrusion detection systems (IDS) are critical for safeguarding computer networks by identifying malicious activities. However, distinguishing attacks in IDSs with high accuracy is challenging. This research proposes a novel approach to enhance the accuracy of anomaly-based intrusion detection systems (IDS). This approach involves combining the Local outlier factor (LOF) algorithm for outlier detection and the Convolutional neural network (CNN) for classification. Firstly, the LOF algorithm is employed to evaluate the local density of network traffic instances, facilitating the identification of outliers deviating significantly from their neighboring data points. Subsequently, a CNN model is utilized for the classification of network traffic instances, effectively categorizing normal and abnormal behavior. CNN's strength lies in its ability to automatically extract relevant features from network traffic data through convolutional layers, thereby enhancing classification performance. The proposed approach achieves 99.87% accuracy in detecting and classifying anomalies in the public dataset of CSE-CIC-IDS2018. This remarkable result underscores the effectiveness of the combined LOF and CNN approach in accurately identifying malicious activities while minimizing false positives. The proposed approach offers valuable insights for researchers and practitioners in the field of network security, empowering them to develop more robust and effective intrusion detection systems.