Abstract
Security risks in the network grow with increase in size. The attacks on networks have increased in recent past tremendously and require efficient intrusion detection systems. Data mining have been used for intrusion detection and have gained much popularity. This paper presents a novel approach to classify intrusion attacks. The central idea is to apply alternating decision trees (ADT) to intrusion data and classify the various types of attacks. Alternating decision tree is well known decision tree algorithm used for binary classification problems. The ADT creates a DT comprised of prediction nodes and splitter nodes. The ADT algorithm is a supervised boosting algorithm. This paper focuses to classify attacks using ADT.The NSL-KDD Data set is used for our experimental analysis. Our proposed method obtained an accuracy of 97.61%, 97.15%, 97.15% and FAR of 3.3, 5.5, 2.38 for DOS, Probe and U2R and R2L respectively, which is much higher than other existing approaches.