Conventional networks were formerly used to transfer data between nodes. The main issue with these networks was that they weren't particularly dependable and couldn't accommodate freshly added devices. As a result, conventional networks are being replaced with SDNs (Software Defined Networks) which carry data of multiple networking applications. In essence, SDNs are dynamic and can be used as foundations for applications that need a lot of data like big data. Centralizations of SDNs are their major advantages. They distinguish switches, routers, and other elements in their environments while forwarding packet. They separate control planed from user/ data planes where the controllers are essentially server-based programmes that instruct switches or routers on how to route data packets. Open Flows are essential parts of SDN installations. They have a set of protocols for direct communications with controllers residing in control planes across SDNs. Flow controls are managed using APIs (Application Programming Interfaces) are employed. Controllers govern traffic by abiding to networking rules. One major problems envisioned in SDNs is the functionality of APIs in terms of data security making it imperative to prioritise network data security in these networks. Network data can be categorised using AIs (Artificial Intelligences) in order to identify malicious or invasive packets. To find abnormalities in SDNs, several MLTs (Machine Learning Techniques) have been employed by researches including DTs (Decision Trees), RFs (Random forests), J48, NBs (Naive Bayes) and DLTs (Deep Learning Techniques). This research work suggests AIMPD (Artificial Intelligence based Malicious Packet Detections) schema based on OANNs (Optimised Artificial Neural Networks) for classifying malicious packets early from SDN traffic information using SDN traffic dataset. The schema showed classification accuracy of above 95% which is evaluated in terms of training and validation accuracy and loss.