AI-Driven Detection of Phishing Attacks through Multimodal Analysis of Content and Design
Shoeb Ali Syed · International Journal of Innovative Research in Computer and Communication Engineering · 2024
Machine operators continue facing dangerous phishing attacks through text manipulation and deceptive visual design schemes that trick users. The prevalent detection procedures of traditional systems depend on static rulebased filters and content-only analysis because these methods fail to keep pace with contemporary phishing methods. The suggested AI-based solution for phishing detection utilizes a dual method that examines written content while assessing webpage designs to achieve better results and flexible system operations. The research employs a mix of legitimate and phishing emails and webpages to implement NLP text features alongside CNN network analyses of visual design elements. The system evaluates linguistic urgency signals, deception markers, and design features, including layout inconsistency, corrupted logos, and color slips. Multimodal analysis outclasses single-modality assessment in experimental tests because it produces superior detection results with higher precision and recall metrics, and F1 scores against phishing attempts. The model can detect new phishing samples that it has not encountered before testing. The study demonstrates the value of semantic and visual data in cybersecurity systems because dynamic and intelligent models are required to fight advanced phishing attacks. This proposed framework provides operational benefits for current phishing defense systems operating in email platforms, web browsers, and enterprise security systems.