As the complexity and scale of present day computer networks proceed to extend, the challenge of detecting anomalous activities within these networks becomes increasingly critical for ensuring cybersecurity. Traditional methods of network anomaly detection often struggle to adapt to the dynamic and advanced nature of modern cyber threats. This project aims to implement deep learning models for detecting several attacks and classifying them based on learning complex patterns using rare anomalies detected from the traffic data. Many feature selection and re-sampling methods will also be tested to get the highest accuracy possible. the chosen models will be tested on UNSW-NB15 and KDD-CUP99 in order to cover as many types of attacks and protocols as possible and to be ready to be implemented later.

• The main aim for the project is to developing a System that can effectively identify and mitigate network anomalies.

• The System will start with emphasis on detecting the 9 types of attacks, with the potential for further scalability.

• The system will utilize suitable feature selection and re-sampling algorithms to enhance its detection capabilities.

• The key objective of this project is to achieve high accuracy in detecting network anomalies through the effective implementation of our models.

• A further goal is to minimize the rates of false positives and false negatives, thereby enhancing the reliability and efficiency of the IDS.

• Our project is mainly designed to detect the anomaly network activities and to identify them to normal and abnormal with classification.

• Resampling: By using resampling methods (SMOTE, ADASYN) to balance out our datasets, then compare between the methods to identify the best one with the best results.

• Feature selection: By using these methods (correlation, random forest, information gain, selectbest, igrf, rfe, chi2) , this process not only improves model accuracy by eliminating redundant or irrelevant data that can lead to overfitting but also reduces computational complexity, leading to faster and more efficient model training.

• Classify into Several different types of attacks.

• Implementation: Implement a software which will Visualize the model and it produces graphs and diagram for the use to make the output understandable.