Nowadays, every individual is interchange the data from information systems, that are more open to the Internet and communication medium, the value of security of networks is extremely increased because of its tremendous utilization. In Internet different types of server are connected to each other now they are under the threats of network attackers. Actually, Intrusion Detection System (IDS) is the second level of defense for which it can be the most powerful system that handles the Attacks done at computer System by making alerts to do the analysts take some sort of actions to prevent this Attacks. IDS are based on the Principle of that an attacker behavior will be clearly different from that of a genuine user. In the proposed system we use a KDD-NSL dataset which will be as the first line of implementation for collect different attribute related to network packet then extract certain attributes from the actual dataset and use such attribute parameter is used to make training dataset and save it into the database. Our training dataset includes 4500+ data rows of values and forty-one attributes. Then in next step is to implement a real-time IDS again find out the different network packets features from dataset according attribute then load training dataset then apply artificial neural network algorithm which is work in three layers input layer, output layer and hidden layer which is a Back Propagation (BPN) and Feed Forward (FF) algorithms so that it provides two outputs normal packets and attacks packet. Proposed system evaluated on the base of performance are classified correctly for both anomaly-based detection and misuse based detection using a dataset of network packets and normal packets.