Artificial intelligence enhanced hybrid intrusion detection model for local area networks

In today's digital environment, cyber security threats are evolving rapidly and it is becoming more difficult to provide effective protection against these threats. Intrusion Detection Systems (IDS) are recognized as one of the key elements of modern information security strategies. Intrusion detection performance of IDS systems is continuously improved with the help of artificial intelligence models, but the desired intrusion detection performance has not yet been achieved. In this study, a data set has been created for the detection of cyber attacks on the local networks of organizations. A dataset was created using Arp Spoofing, MAC flooding, DHCP starvation, STP Root Bridge and Rogue DHCP server attack vectors, which are frequently encountered in local networks and cause data loss and service interruptions. Extreme Gradient Boost (XGBoost) classifier, a gradient boosting algorithm based on decision trees, was used to measure the intrusion detection capacity of the generated dataset. As a result of the hyperparameter optimization to improve the intrusion detection performance, an accuracy of 89.76% was achieved on the proposed dataset. In addition, the attack detection time, which is among the most crucial intrusion detection criteria, is as short as 8.32 seconds for the suggested model. These outcomes demonstrate the effectiveness of the suggested methodology and dataset in identifying intrusions in limited networks.