Review on Phishing Threat Advancement in Todays Digital Environments: Detection Techniques, Agents and Future Directions

Abstract

Phishing threats serve as a continuous advanced digital security threat which remains active throughout all contemporary online platforms. The researchers conducted a systematic review to study the historical development of phishing link attacks by analyzing 26 peer-reviewed studies that researchers published between 2020 and 2025. The review employs the PRISMA 2020 framework to evaluate phishing agents and attack strategies and detection methods and upcoming research paths which it establishes through evidence from IEEE and ACM and ScienceDirect and Scopus databases. The research shows artificial intelligence combined with social engineering methods and multi-channel delivery systems which include email and SMS and voice calls and social media platforms results in more sophisticated phishing attacks. The research establishes machine learning and deep learning techniques which include CNNs and RNNs and transformers and GCNs and multimodal models as the most effective methods to identify phishing URLs and emails because they achieve more than 95% accuracy across different research studies. The deployment of machine learning solutions encounters difficulties because adversarial attacks and domain adaptation issues and the need for standardized benchmarking datasets remain unresolved. The review establishes that future defense systems need to incorporate explainable AI together with adversarial-resilient models and cross-dataset standardization and multimodal architectures and human-centric training strategies for effective defense against upcoming phishing threats.