CYBERSECURITY ENHANCEMENT THROUGH DEEP LEARNING AND ANOMALY DETECTION FOR CYBER THREAT DETECTION

Abstract

The rapid advancement of digital technologies has made cybersecurity essential for safeguarding computer systems and networks. It involves techniques and algorithms to prevent unauthorized access, data breaches, and malicious activities while ensuring data security and privacy. A multi-layered defense on the host, application, network, and data levels is essential and continually enhanced to suit challenging cyber threats. With attack patterns getting increasingly advanced, cybersecurity has become a core element that supports trust and resilience in the digital age. This paper used the benchmark NSL-KDD data to consolidate an efficient cyber threat detection system. Data preparation includes filling in missing values, removing duplicates, cleaning up noise, one-hot encoding, normalizing, and using Random Forest to improve feature selection for a better model performance. Training and testing were conducted using deep learning and machine learning techniques such as ANNs, SVMs, DTs, and RNNs (Recurrent Neural Networks). At 96% accuracy, 92% precision, 93% recall, and 92% F1-score, RNN was the most successful model out of all that were evaluated. The findings indicate that combining feature selection and innovative deep learning methods remarkably enhances the detection accuracy, robustness and as such supports the increase in cybersecurity and securing against emerging cyber threats.