Supervised Machine Learning based Routing Detection for Smart Meter Network

It is known that the Ad hoc On-Demand Distance Vector (AODV) routing protocol for smart meter network is vulnerable to denial of service attacks (e.g., black hole attack and selective forwarding attack). In this paper, we introduce supervised machine learning to detect unknown routing attacks under AODV. There are two problems in the existing intrusion detection algorithms. The first problem is that the existing intrusion detection algorithms are mainly applied to a specific and known type of routing attack, which no longer work for unknown attacks. The second one is that constant thresholds are commonly used for detection. To overcome these two problems, we introduce a supervised machine learning based detection approach. To implement supervised machine learning, three steps are involved. First, features and target estimations are selected from malicious AODV behaviors in smart meter network to generate training data sets. Second, we assign a suitable classifier including support vector machine, k-nearest neighbors and decision trees to t the training and predicted data. Third, we update our training data to maintain a dynamic threshold. Simulations are conducted using Python3.6 to evaluate the accuracy and the time overhead of our pro- posed supervised machine learning model. The simulation results show that the decision trees algorithm assures 100% accuracy with minimum time overhead to detect routing attacks in AODV.