A Hybrid Approach to Intrusion Detection: Autoencoders and Decision Trees for Cybersecurity

Intrusion Detection Systems (IDS) are vital for detecting and mitigating malicious network activities, but traditional models struggle with overfitting, feature selection, and computational inefficiencies, limiting their adaptability to evolving cyber threats. This study addresses these challenges by developing a hybrid IDS framework leveraging autoencoders for feature reduction and decision trees for classification. Using the CSE-CIC-IDS2018 dataset, the research incorporates robust preprocessing, balanced data sampling, and iterative experimentation to optimize model performance. The proposed framework enhances IDS efficiency, scalability, and adaptability, offering a practical solution for real-world cybersecurity challenges.