The rapid expansion of the internet has resulted in an increase in cyber threats, [1, 2] particularly through malicious URLs that are widely used in phishing attacks, malware distribution, and fraudulent activities. Conventional methods using blacklists do not manage well against increasing dangers. This study proposes a new URL-based cyber-attack detection system that uses deep learning with feature extraction. Several deep learning models were developed and tested, and the one which performed the best achieved very high precision and recall ratios. The system provides immediate classification and is able to classify URLs as either pages and or websites, as well as determine whether they are harmful or not. The proposed model improves the level of cybersecurity and threat detection efforts.