SummaryDue to the widespread distribution of network connectivity, the demand for network security and protection against cyber‐attacks is increasing rapidly. Intrusion detection system (IDS) plays a vital role in the network security. Previously, the concept‐drift based IDS is performed, in which the suspicious activity at specific IP address is detected on the real time network. However, in that research the intrusion detection is performed on single attribute. In this work, an efficient adaptive window with support vector machine (ADWIN‐SVM) is developed to detect the intrusions on large datasets with multiple attributes. The initial process in the proposed system is data cleaning. To clean the input data, the data normalization technique is utilized. Then, the data are streamed as online data by including the timestamp attribute to the dataset. This process assists to extract the patterns and detect the changes in the data. ADWIN algorithm identifies the sudden generation of drift and separates the window based on the split point. The data can be classified as normal and attack by the SVM classifier. For the performance evaluation of the proposed IDS, two datasets such as KDD Cup 1999 and CICIDS 2017 are utilized. Python is the implementation tool and the performance metrics considered for the proposed approach are precision, recall, F1 score, accuracy, and false positive rate. The proposed ADWIN‐SVM performance is compared with various existing techniques such as random forest, decision tree, naive Bayes, logistic regression, AdaBoost, K‐nearest neighbor, and SVM‐rbf. From the comparative analysis, the proposed ADWIN‐SVM is outperformed in intrusion detection.