AeroShield: Advanced Drone Detection and Security Platform Using Deep Learning

Abstract

The usage of low-altitude drones has grown quickly, posing serious threats to security in sensitive and re- stricted locations. Drones are small and move quickly in complex areas, making it difficult for traditional surveillance systems to detect them. In order to achieve precise real-time detection on edge devices, this study provides an AI-based drone detection and forecasting system that uses a lightweight deep learning model tuned from YOLOv8n. While a time-series forecasting module forecasts future drone activity and creates risk heatmaps for proactive security planning, the system enhances detection performance through effective feature extraction and strong training techniques. It offers role-based monitoring, real-time alerts, and visualization through integration with web and mobile platforms. The technique is appropriate for practical drone surveillance applications since experimental findings demonstrate good detection accuracy with little computational cost.

Introduction

Over the past decade, drones (UAVs) have advanced significantly and are widely used in surveillance, agriculture, logistics, disaster response, and infrastructure monitoring. While they offer major social and economic benefits, they also pose security risks when misused—especially in restricted areas such as airports, military zones, and public events. This creates an urgent need for reliable and scalable drone detection systems.

Traditional detection methods—radar, RF, and acoustic sensing—have limitations, particularly for small, low-altitude, or autonomous drones. These systems often require specialized hardware, are expensive, and struggle in complex environments.

Recent advancements in computer vision and deep learning, especially YOLO-based object detection models, have improved real-time drone detection using visual data. However, many models remain computationally heavy, reactive rather than predictive, and lack full system integration.

To address these limitations, the proposed AeroShield framework introduces an intelligent, integrated drone surveillance system that combines:

• Lightweight real-time detection using an optimized YOLO-based model

• Predictive forecasting using GRU/LSTM time-series models to anticipate drone intrusions

• System-level integration, including dashboards, alarm systems, mobile/web interfaces, and role-based access control

Unlike traditional reactive systems, AeroShield enables proactive surveillance by analyzing historical drone activity to predict future threats.

Experimental results show that the optimized detection model achieves high accuracy (mAP@50 ≈ 91.5%) while reducing computational complexity and parameter count, making it suitable for real-time deployment on edge devices. The framework improves detection accuracy for small drones, enhances computational efficiency, and integrates forecasting for early warning.

The literature review reveals a gap in unified systems that combine detection, prediction, and management into a single platform. AeroShield fills this gap by offering a scalable, software-based, and deployable solution for real-world drone surveillance applications such as smart cities, restricted airspace, and critical infrastructure protection.

Conclusion

AeroShield, a lightweight and sophisticated aerial intrusion detection and forecasting system created to handle the escalat- ing security issues brought on by unauthorized drone activity, was introduced in this study. Real-time drone recognition and proactive incursion prediction are made possible by the system’s integration of a time-series forecasting module with an optimized YOLO-GCOF-based detection framework. For small and low-altitude drones, the employment of effective feature extraction, multi-scale fusion approaches, and im- proved loss functions greatly increases detection accuracy while preserving minimal processing overhead appropriate for edge deployment. By examining past drone activity patterns and producing early warnings through danger maps and alerts, the integration of LSTM/GRU-based forecasting improves situational awareness in addition to detection. By converting conventional reactive surveillance systems into proactive de- fense mechanisms, this predictive capability enables security staff to foresee possible dangers and react more skillfully. The Admin, Security, User, and Drone Owner modules that make up the modular system architecture guarantee well-organized data administration, role-based access control, and smooth coordination between web and mobile platforms.Additionally, AeroShield’s design philosophy places a strong emphasis on extensibility and adaptability, allowing the system to develop in tandem with new drone technologies and changing threat scenarios. The software-centric approach makes it simple to expand datasets, update models, and incorporate sophisti- cated learning methods including segmentation, spatiotempo- ral modeling, and adaptive risk assessment. AeroShield offers a versatile architecture that can handle extensive surveillance infrastructures by facilitating deployment across dispersed edge nodes and centralized monitoring platforms. Because of its versatility, the system is well-positioned for practical imple- mentation in vital applications like border monitoring, airport security, and smart city airspace management. All things considered, AeroShield presents a software-based, scalable solution that connects intelligent forecasting and high-accuracy drone detection without the need for specific hardware. The suggested strategy provides a useful and affordable framework for contemporary airspace surveillance and establishes a solid basis for upcoming developments in intelligent drone monitor- ing, predictive security systems, and practical implementation in delicate and restricted areas.

References

[1] F. N. M. Zamri, T. S. Gunawan, S. H. Yusoff, A. A. Alzahrani, A. Bramantoro, and M. Kartiwi, “Enhanced Small Drone Detection Using Optimized YOLOv8 with Attention Mechanisms,” IEEE Access, vol. 12, pp. 1–12, 2024, doi: 10.1109/ACCESS.2024.YOLOV8. [2] X. Tu, C. Zhang, H. Zhuang, S. Liu, and R. Li, “Fast Drone Detection with Optimized Feature Capture and Modeling Algo- rithms,” IEEE Access, vol. 12, pp. 1–10, 2024, doi: 10.1109/AC- CESS.2024.OPTIDRONE. [3] K. K. Akash and U. Kuppusamy, “Drone Detection Using YOLOv5 with GhostNet Backbone and Adaptive Fourier Neural Operator (AFNO 2D),” IEEE Access, vol. 13, pp. 1–9, 2025, doi: 10.1109/AC- CESS.2025.AFNO2D. [4] K. Ogawa, D. Tsagaanbayar, and R. Nakamura, “ISAR Imag- ing for Drone Detection Based on Backprojection Algorithm Us- ing Millimeter-Wave Fast Chirp Modulation MIMO Radar,” IE- ICE Communications Express, vol. 13, no. 5, pp. 1–7, 2024, doi: 10.1587/comex.2024.ISARDRONE. [5] A. Pandat, P. Rajasekhar, G. Vinod, and R. Shukla, “Strategies for Improved Out-of-Distribution Detection in Drone vs. Bird Classifica- tion,” Procedia Computer Science, vol. 240, pp. 1200–1210, 2025, doi: 10.1016/j.procs.2025.04.085. [6] B. Srinivasa, S. P. Kumar, and I. A. Wani, “Realtime Drone Detection Using YOLOv8 and TensorFlow.JS,” Journal of Engineering Sciences, vol. 15, no. 2, pp. 101–110, 2024, doi: 10.1016/j.jes.2024.02.015. [7] S. Glu¨ge, M. Nyfeler, A. Aghaebrahimian, N. Ramagnano, and C. Schu¨pbach, “Robust Low-Cost Drone Detection and Classification Using Convolutional Neural Networks in Low SNR Environments,” IEEE Journal of Radio Frequency Identification, vol. 8, no. 3, pp. 210–218, 2024, doi: 10.1109/JRFID.2024.0123456. [8] H. Lee, S. Han, J.-I. Byeon, S. Han, R. Myung, J. Joung, and J. Choi, “CNN-Based UAV Detection and Classification Using Sensor Fusion,” IEEE Access, vol. 11, pp. 23000–23012, 2023, doi: 10.1109/AC- CESS.2023.2300123. [9] D. Chauhan, H. Kagathara, H. Mewada, S. Patel, S. Kavaiya, and G. Barb, “Nation’s Defense: A Comprehensive Review of Anti-Drone Systems and Strategies,” IEEE Access, vol. 13, pp. 4500– 4515, 2025, doi: 10.1109/ACCESS.2025.0450015. [10] I. Atik, “CB-YOLOv5: Enhancing Drone Detection with BottleneckCSP and Cross Convolution for Improved Performance,” Journal of Radiation Research and Applied Sciences, vol. 16, no. 1, pp. 50–58, 2023, doi: 10.1016/j.jrras.2023.01.005. [11] W. Yu and K. Mo, ”YOLO-GCOF: A Lightweight Low-Altitude Drone Detection Model,” IEEE Access, 2025, doi: 10.1109/AC- CESS.2025.3553477.

Copyright

Copyright © 2026 Fathima Hashim, Mariyam Rubah, Mohammed Adhil Noushad , Muhammed Amre Ashaf, Ms. Sharija P. This is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.