AI-DRIVEN SOCIAL ENGINEERING ATTACK DETECTION IN EMAIL COMMUNICATIONS

The goal of this research is to create an efficient AI driven framework to detect social engineering attacks in email communications that have been heavily leveraged to exploit human vulnerabilities and which are on the rise.The study leverages state of the art Natural Language Processing (NLP) techniques and deep learning method over hybrid designed Bidirectional Encoder Representations from Transformers (BERT) and Long Short Term Memory (LSTM) architectures.The technique is to gather and preprocess publically available datasets of malicious and benign email samples and use semantic and behavioral analysis to extract features.The experimental results show that the proposed model can well classify deceptive emails with 97.3% accuracy which is better than traditional machine learning algorithms.What is particularly interesting is that the model is able to detect very minute linguistic patterns indicating emotional manipulation, that are mostly missed by standard filters.It also emphasizes on the role of emerging technologies like transformer based models and context aware NLP techniques in the cybersecurity domain to strengthen the resilience against the targeted human factor vulnerabilities.The proposed system, through proactive threat detection and response, plays a big role in improving email security frameworks in corporate and individual environments and thus bolsters cyber resilience against the ever changing social engineering threats.