Social Engineering Attack Classifications on Social Media Using Deep燣earning

In defense-in-depth, humans have always been the weakest link in cybersecurity. However, unlike common threats, social engineering poses vulnerabilities not directly quantifiable penetration testing. Most skilled engineers trick users into giving up information voluntarily through attacks like phishing and adware. Social Engineering (SE) media is structurally similar to regular posts but contains malicious intrinsic meaning within sentence semantic. this paper, a novel SE model trained using Recurrent Neural Network Long Short Term Memory (RNN-LSTM) identify well-disguised threats posts. We use custom dataset crawled from hundreds of corporate personal Facebook First, attack detection pipeline (SEAD) designed filter out with intents domain heuristics. Next, each post tokenized sentences then analyzed sentiment analyzer before being labelled as an anomaly or normal training data. Then, we train RNN-LSTM detect five types that potentially contain signs gathering. The experimental result showed Attack (SEA) achieves 0.84 classification precision 0.81 recall compared ground truth labeled by network experts. results semantics linguistics similarities are effective indicator for early SEA.