TIDCS: A Dynamic Intrusion Detection and Classification System Based Feature Selection

摘要: Machine learning techniques are becoming mainstream in intrusion detection systems as they allow real-time response and have the ability to learn adapt. By using a comprehensive dataset with multiple attack types, well-trained model can be created improve anomaly performance. However, high dimensional data present significant challenge for machine techniques. Processing similar features that provide redundant information increases computational time, which is critical problem especially users constrained resources (battery, energy). In this paper, we propose two models classification scheme Trust-based Intrusion Detection Classification System (TIDCS) System- Accelerated (TIDCS-A) secure network. TIDCS reduces number of input based on new algorithm feature selection. Initially, grouped randomly increase probability making them participating generation different groups, sorted their accuracy scores. Only ranked then selected obtain any received packet from nodes network, saved part node’s past proposes periodic system cleansing where trust relationships between participant evaluated renewed periodically. TIDCS-A dynamic compute exact time states restricts exposure window nodes. The final decision both estimated by incorporating behavior algorithm. Any detected trustworthiness involved, leading cleansing. An evaluation NSL-KDD UNSW datasets shows detect malicious behaviors providing higher accuracy, rates, lower false alarm than state-of-art For instance, dataset, 91% TICDS, 83.47%by online AODE, 88% CADF, 90% EDM, TANN 69.6% NB. Consequently, TICDS has better performance state art terms detection, while good rates.