A robust ensemble of neuro-fuzzy classifiers for DDoS attack detection
Ali Sharifi Boroujerdi, Saeed Ayat · 2013
Recent studies show that Distributed Denial of Service (DDoS) attacks play an important role in the security of computers because they can decrease the efficiency of victim resources within a short period of time. In this paper, an innovative ensemble of Sugeno type adaptive neuro-fuzzy classifiers has been proposed for attack detection using an effective boosting technique named Marliboost. Detection accuracy and false positive alarms are two measurements used to evaluate the performance of the proposed technique. Experimental results on the optimized randomly selected subset of NSL-KDD confirm that the proposed ensemble of classifiers has higher detection accuracy (96%) in comparison with the other widely used machine learning techniques. Moreover, false positive alarms have been greatly reduced by applying the presented technique.