Abstract: The significant increase in computer network usage and the huge amount of sensitive data being stored and transferred through them has escalated the attacks and invasions on these networks. Secure data communication over the internet and any other network is always under threat of intrusions and misuses. The system that monitors the events occurring in a computer system or a network and analyzes the events for signs of intrusion is known as an Intrusion Detection System (IDS). In information protection, the Intrusion Detection System (IDS) has become a crucial component in terms of computer and network security which monitors the network traffic to detect possible security threats. There are various approaches being utilized in intrusion detections but unfortunately any of the systems so far are not completely flawless and suffer from a number of drawbacks such as low accuracy to detect new types of intrusions and misclassification of normal and malicious traffic, in addition to long response time. It is necessary to develop an IDS that is accurate, adaptive and extensible to overcome these weaknesses. In this study, we proposed a learning-based method which improves IDS adaptability to new attacks and reduces false alarms. The method that has a distributed architecture to increase performance and scalability of the IDS and uses C4.5 decision trees with the feedback learning technique to adapt dynamic network behaviors. To evaluate the proposed method we used some well-known datasets in this context such as KDD Cup 99 and did several tests with approximately 97% detection accuracy on benchmarks. According to the promising results, the adaptable IDS approach is more accurate than traditional systems and it is more efficient against new complex network attacks.