Predicting ARP spoofing with Machine Learning

Securing a network from potential attacks has become a great challenge after the advent of sensor networks that consists of different sensors connected by wireless settings, to sense and share the collected data. Such networks can be compromised by different attacks including and not limited to Denial of Services (DOS), Distributed DOS, and Man-in-the-Middle types of attacks. Detecting these attacks early is a challenging task. Like many other fields, Machine Learning and Deep Learning can be utilized here to predict the attack early. This research uses Long Short-Term Memory (LSTM) networks and decision tree classifiers to predict ARP Spoofing. The performance of the models is evaluated on a comprehensive dataset that contains data on different sorts of attacks. Results of different experiments showed that both can predict the intrusion quite well. The accuracy achieved by LSTM and decision tree was 99% and 100% respectively which is better than other state-of-art solutions. The decision tree beats the complicated LSTM network on execution speed.