Abstract
Intrusions are critical issues in information system of healthcare sector because a sole intrusion can cause health issue due to any manipulation in the medical record of the patients. Several intrusion detection (ID) techniques have been used but their performance is the dilemma. The efficiency of intrusion detection systems (IDSs) depends on optimal classifier architecture to categorize the data into intrusive or normal, which required increasing detection rates (DR) and decreasing false alarm rates (FAR). Therefore, to find an optimal classifier architecture to enhance performance in IDSs is an important subject. This study proposed Elman Neural Network-based IDS as classification technique in order to enhance performance. NSL-KDD Dataset is used for evaluation and assessment. Moreover, Principle Component Analysis (PCA) is applied in this work in order to convert raw features to principal space and choose features based on their sensitivity. The proposed approach is capable to enhance performance by increased DR and decreased FAR.