In recent years, balloon-borne threats carrying hazardous or explosive materials have emerged as a novel form of asymmetric terrorism, posing serious challenges to public safety. In response to this evolving threat, this study presents an AI-driven autonomous drone defense system capable of real-time detection, tracking, and neutralization of airborne hazards. The proposed framework integrates state-of-the-art deep learning models, including YOLO (You Only Look Once) for fast and accurate object detection, and convolutional neural networks (CNNs) for X-ray image analysis, enabling precise identification of hazardous payloads. This multi-stage system ensures safe interception and retrieval while minimizing the risk of secondary damage from debris dispersion. Moreover, a robust data collection and storage architecture supports continuous model improvement, ensuring scalability and adaptability for future counter-terrorism operations. As balloon-based threats represent a new and unconventional security risk, this research offers a practical and deployable solution. Beyond immediate applicability, the system also provides a foundational platform for the development of next-generation autonomous security infrastructures in both civilian and defense contexts.

With the rapid acceleration of the Fourth Industrial Revolution, cyber threats have become increasingly frequent and complex. However, most public and private institutions still rely heavily on manual cybersecurity operations, which often lead to delayed responses and human errors, exposing critical vulnerabilities. A particular challenge lies in the inability to efficiently integrate and automate the analysis of threat logs collected from various sources, limiting the effectiveness of threat prediction and mitigation. To address these challenges, this study proposes an AI-based RPA (Robotic Process Automation) system designed to automate the collection, analysis, and dissemination of cyber threat logs. By minimizing human intervention, the proposed system significantly enhances real-time response capabilities and reduces errors. Additionally, standardizing and centralizing diverse log formats lays a foundation for the future development of AI models capable of predicting cyberattack patterns. This system is particularly well-suited for government and public organizations, offering a cost-effective solution that enhances cybersecurity while maintaining compatibility with existing infrastructures. The experimental results demonstrate that the proposed AI-based RPA system outperforms traditional manual systems in terms of log processing speed, prediction accuracy, and error reduction. This study highlights the critical role of automated AI-driven systems in enabling real-time threat response and prevention, presenting a practical and scalable approach for modern cybersecurity environments.

During the golden-time phase of urban disasters, emergency response systems must issue timely directives that remain compliant with jurisdictional policies and regulations, even when early human reports precede reliable CPS telemetry and network connectivity is degraded. This paper presents a policy-grounded, multilingual natural-language control layer for public safety cyber-physical systems (CPS), designed to operate under strict latency, accountability, and connectivity constraints. The proposed system converts short, unstructured incident reports and CPS state summaries into role-specific, CPS-ready directives by integrating three components: (i) clause-level retrieval from a fixed and locally verified policy snapshot, (ii) grounded reasoning using a lightweight local language model, and (iii) deterministic release-time safety validation that enforces regulatory constraints and hazard-specific prohibitions. All inference-time operations are self-contained and support offline execution.We evaluate the system across English, Japanese, and Korean using 3,600 executed trial sets (1,200 per language), where each trial set is a five-query bundle spanning five hazard categories executed sequentially under identical deterministic settings. Under both connectivity-available and offline operation, the proposed engine achieves a median end-to-end latency of 13.8 s per trial set while maintaining 95.0% clause-level policy traceability on released directives and substantially suppressing unsafe releases under adversarially phrased inputs. These results demonstrate that deterministic policy grounding and offline-capable local inference can serve as a deployable and auditable control layer for CPS-assisted disaster response under golden-time constraints.