Abstract:
Innovative security solutions utilizing cutting-edge machine learning techniques are essential to strengthen network defense
as cyber threats become more sophisticated. This study proposes an intrusion detection system (IDS) that uses deep learning algorithms
(DLAs), specifically convolutional neural networks (CNNs) and, to automatically detect phishing attacks. Phishing circumvents
traditional signature-based intrusion detection systems by using cunning social engineering techniques. CNNs enable the automatic
extraction of sophisticated features from raw input data, including URLs and webpage content. Low-level patterns are recognized by
their convolutional layers, and in later layers, these patterns are combined to create higher-level representations. Sequential data, such
as user activities over time, is a great fit for CNN modeling. CNNs work together to acquire the intricate multi-modal patterns that are
characteristic of phishing. Back propagation-based model optimization enables real-time adaptation to identify emerging phishing
variants. DLA integration with an IDS offers a strong defense against sophisticated user-targeted phishing attacks. Using the KDDCUP99
dataset, which has 175,341 training and 82,332 testing instances, the DLA model is trained. Proactive incident response is
made possible by automated feature learning by DLAs, which dramatically increases detection accuracy over manual rule-based
techniques. This DLA-driven intrusion detection system research strengthens the overall security posture by improving resistance to
changing social engineering threats. By utilizing machine learning, networks and users can be protected from sneaky phishing tactics
through constant model refining for intelligent, adaptive threat identification as attack vectors change. The accuracy of the Phishing
Detection with an accuracy of 99.2% and with a Model Loss of 79%.
