<![CDATA[The increasing number of phishing and social engineering attacks through email has made it more difficult for users to distinguish between legitimate and potentially harmful emails. Detection approaches based solely on email authentication or content analysis still have limitations in understanding message context and validating the sender. This study aims to develop an early warning system for detecting indications of social engineering in emails using a hybrid decision approach. The research employed the Prototype method by combining email authentication analysis through SPF, DKIM, DMARC, and domain alignment with content analysis using the Qwen2.5:1.5B Small Language Model running locally through Ollama. The evaluation was conducted using 112 email samples consisting of 63 legitimate emails and 49 emails containing indications of social engineering. The proposed mechanism achieved an accuracy of 94.64%, precision of 95.74%, recall of 91.84%, and F1-score of 93.75%. These results indicate that the hybrid decision approach improves email detection capability in a more adaptive manner. Therefore, the proposed mechanism can support the implementation of an early warning system against email-based social engineering threats.]]>
Copyrights © 2026
