Insider Threat Surveillance System via Multimodal-Based Physical Leak Detection and Optimal Imbalanced Augmentation in Research and Development Environments

초록

Increasing security incidents in research and development (R&D) demand enhanced measures to prevent technical leaks, especially against insider threats. Despite existing guidelines to mitigate external threats, current measures often overlook physical leaks in open collaborative environments. To the best of our knowledge, this study is the first to propose a surveillance system for detecting insider threats arising from physical paths. Security requirements are derived from a physical security perspective, and an audiovisual dataset is constructed. Additionally, this study presents a new augmentation technique to address overfitting by establishing an optimal point for augmenting imbalanced leakage versus normal behavior based on the proposed formula. Consequently, the proposed method could achieve 91.26% accuracy in leakage detection and 99% in distinguishing eight types of single events. This result validates the effectiveness of the proposed dataset and method in R&D settings, establishes a new academic standard for physical leakage detection, and introduces a universal augmentation formula for multiclass classification. In conclusion, this method enhances industrial surveillance systems and mitigates insider threats in R&D environments. Future studies should focus on developing a leakage behavior dataset based on big data using multiple criteria and building a real-time surveillance system suitable for industrial environments. © This is an Open Access article distributed under the terms of the Creative Commons Attribution Non-Commercial License (http://creativecommons.org/licenses/by-nc/3.0/) which permits unrestricted non-commercial use, distribution, and reproduction in any medium, provided the original work is properly cited.

키워드