<![CDATA[The digital transformation of educational assessment systems has accelerated the adoption of computer-based technologies; however, it still faces significant challenges related to security and identity verification of examination participants. One of the major issues is impersonation, where unauthorized individuals act as proxies during exams, thereby compromising academic integrity. This study aims to develop and evaluate SisCek (Sistem Pendeteksi Calo Ujian/ Exam Broker Detection System) based on face recognition and deep learning as a solution to automatically and in real time detect and prevent such practices. The research employs an experimental approach involving facial data collection, preprocessing, model training using a Convolutional Neural Network (CNN), and integrated system implementation. The evaluation is conducted using accuracy, False Acceptance Rate (FAR), and False Rejection Rate (FRR), as well as testing under real examination scenarios. The results show that the proposed model achieves an accuracy of 96.8%, with a FAR of 2.1% and an FRR of 3.4%. System-level testing demonstrates a detection success rate of 96% for both legitimate participants and impostors, with an average response time of 2.5 seconds, satisfying real-time system requirements. Comparative analysis indicates that SisCek outperforms conventional systems and previous studies, particularly in real-time impersonation detection and full integration with examination systems. This study provides a significant contribution to the development of AI-based examination security systems and has strong potential to enhance the integrity, fairness, and credibility of educational assessment in the digital era.]]>
