Building strong IDS is essential in today's network traffic environment, feature reduction is one approach in constructing the effective IDS system by selecting the most relevant features in detecting most known and unknown attacks. In this work, proposing the hybrid feature selection method by combining Mutual Information and Linear Correlation Coefficient techniques (MI-LCC) in producing the most efficient and optimized feature subset. Support Vector Machine (SVM) classification technique being used in accurately classifying the traffic data into normal and malicious records. The proposed framework shall be evaluated with the standard benchmarked datasets including KDD-Cup-99, NSL-KDD, and UNSW-NB15 datasets. The test results, comparison analysis and reference graphs shows that the proposed feature selection model produces optimized and most important features set for classifier to achieve stated accuracy and less false positive rate compared with other similar techniques.