Abstract: Aiming at the diversity and the large amount of monitoring data of MANET (mobile ad hoc networks), an anomaly detection method in MANET based on improved k-means algorithm was proposed. By introducing the classification the contribution degree, the weight of each dimension can be calculated reasonably, and genetic algorithm and k-means were combined to prevent the results of clustering from getting in local optimization. Then, the detection method under the framework of MapReduce was put forward, and parallel clustering was achieved by using population migration strategy . The experimental results show that the detection accuracy and efficiency of the proposed method are better than the traditional ones.

Key words: mobile ad-hoc networks; anomaly intrusion detection; k-means clustering; MapReduce; genetic algorithm; classification contribution degree