Dynamic Machine Learning Algorithm for AODV Routing Attacks Detection

Ad hoc On-Demand Distance Vector (AODV) routing protocol is vulnerable to some routing attacks including blackhole attack and flooding attack. Typically, these two types of routing attacks are linked with two major malicious behaviors: fake Route Replies (RREP) and fake Route Request (RREQ) flooding. In this paper, we develop a novel dynamic machine learning approach to detect blackhole and flooding attacks in AODV. The proposed solution primarily determines three distinct features by analyzing Hello Packet, RREQ and RREP in AODV routing protocol. Then, these features are used to develop a mathematical model for dynamic learning algorithm. After that, we generate the training set of data and assign a threshold for our machine learning model using these features. This training set of data is only valid for N time slots, which is regarded as one iteration. In the following iterations, it will update the latest valid outcomes from the dynamic learning model and determine a updated threshold for the model, which significantly increases the detection accuracy. Extensive simulations have been conducted to evaluate the accuracy and the time overhead of three classifiers, e.g., support vector machine, k-nearest neighbors and decision trees. The simulation results show that the proposed algorithm can achieve very high accuracy with minimum time overhead to detect malicious behavior in AODV routing protocol.