Cyber security has been a major concern due to the large number of cybercrimes that are committed online. These crimes have resulted in large number of exploits and also impairment of significant data, compromising confidentiality and authenticity. Avoiding network exploits aids effectively in curtailing cybercrimes. This work presents a double loop learning ensemble, DLLE that can be used for effective intrusion detection in networked environments. The proposed architecture is composed of two major phases. The initial specialized learning phase uses a one-class learning model for predicting normal and anomalous records. Discrepancies arising in this phase are corrected using the bagging based ensemble that uses double loop learning methods, which is the second phase of the architecture. Experiments were performed using the CSE-CIC 2018 data set. Comparisons were done with state of the art existing models. The results indicate that the DLLE model exhibits 99% accuracy levels and 98.7% precision levels. The performances are indicative of high performance and guarantees highly accurate results when used in real time networked environments

Network Intrusion Detection; Ensemble Learning; Double Loop Learning; Bagging; Rule Based Predictions