A Hybrid Approach to Reliable Jamming Identification in UAV Communications Using Combined DNNs and ML Algorithms

Farkhari H.; Viana J.; Kahvazadeh S.; Sebastião P.; Jimenez V.P.G.; Dinis R.

doi:10.1109/ACCESS.2024.3504729

Published January 1, 2024 | Version v1

Journal article Open

A Hybrid Approach to Reliable Jamming Identification in UAV Communications Using Combined DNNs and ML Algorithms

1. ISCTE-Instituto Universitário de Lisboa, Lisbon, 1649-026, Portugal
2. Universidad Carlos III de Madrid (UC3M), Departamento de Teoría de la Señal y Comunicaciones, Madrid, 28903, Spain
3. CERCA, Centre Tecnològic de Telecomunicacions de Catalunya (CTTC), Barcelona, 08860, Spain
4. ISCTE-Instituto Universitário de Lisboa, Lisbon, 1649-026, Portugal; Instituto de Telecomunicações (IT), Lisbon, 1049-001, Portugal
5. Universidade Nova de Lisboa, Monte da Caparica, FCT, Caparica, 2829-516, Portugal; Instituto de Telecomunicações (IT), Lisbon, 1049-001, Portugal

Deep Neural Networks (DNNs) have gained prominence due to their remarkable accomplishments across various domains, including telecommunications and security. Their integration into decision-making processes within 5G telecommunication systems and UAV security is noteworthy. However, the iterative nature of DNN data processing can introduce uncertainties in classification decisions, impacting their reliability. This paper presents novel combined preprocessing and post-processing techniques designed to enhance the accuracy and reliability of binary classification DNNs by managing uncertainty levels. The study evaluates these methods through calibration error metrics, confidence values, and the Reliability Score (RS), which quantifies the disparity between Mean Accuracy (MA) and Mean Confidence (MC). Additionally, the effectiveness of these methods is demonstrated by applying them to simulated real-world scenarios to improve jamming detection reliability in UAV communications. The proposed algorithms' impact is compared against baseline DNNs and DNNs augmented with the eXtreme Gradient Boosting (XGB) classifier, as well as the latest research to validate our approach. This paper comprehensively overviews the experimental setup, dataset, deep network architecture, preprocessing and post-processing techniques, evaluation metrics, and results. By addressing uncertainty in XGB and DNN outputs, this study improves the trustworthiness of ML-DNN-based decision-making processes in 5G UAV security scenarios.

Notes

Funding text 1: This work was supported in part by the European Union's Horizon 2020 Research and Innovation Programme under the Marie Sklodowska-Curie under Project 813391; in part by the Ministerio de Asuntos Econ\u00F3micos y Transformaci\u00F3n Digital (MINECO), in part by the European Union (EU)-NextGenerationEU in the Frameworks of the \"Plan de Recuperaci\u00F3n, Transformaci\u00F3n y Resiliencia\" and of the \"Mecanismo de Recuperaci\u00F3n y Resiliencia\" under Grant TSI-063000-2021-55 and Grant PID2021-126431OB-I00; in part by the Ministerio de Ciencia, Innovaci\u00F3n y Universidades (MCIN)/Agencia Espa\u00F1ola de Investigaci\u00F3n (AEI)/10.13039/501100011033; in part by the \"European Regional Development Fund (ERDF) A way of making Europe\" and Generalitat de Catalunya under Grant 2021 SGR 00770; in part by Project \"SOFIA-AIR\" PID2023-147305OB-C31, Ministerio de Ciencia, Innovaci\u00F3n y Universidades (MICIU)/AEI/10.13039/501100011033/Ministerio de Ciencia, Innovaci\u00F3n y Universidades (FEDER) EU, in part by FCT-Fund\u0105\u00E3o para a Ci\u00EAncia e Tecnologia, I.P., and Instituto de Telecomunic\u0105\u00F5es under Project UIDB/50008/2020, with DOI identifier https://doi.org/10.54499/UIDB/50008/2020; and in part by the Distributed Access Design for Cell-less Smart 6G Networks (CELL-LESS6G) project, 2022. 08786.PTDC with DOI identifier https://doi.org/10.54499/2022.08786.PTDC.; Funding text 2: \u00A7 Collaborative authors with equal contribution. This research received funding from the European Union\u2019s Horizon 2020 research and innovation programme under the Marie Sklodowska-Curie Project Number 813391. Grant PID2021-126431OB-I00 funded by MCIN/AEI/ 10.13039/501100011033 and by ERDF A way of making Europe. Generalitat de Catalunya grant 2021 SGR 00770. Project \"IRENE\" PID2020-115323RB-C33, MINECO/AEI/FEDER, UE. Funda\u00E7\u00E3o para a Ci\u00EAncia e Tecnologia and Instituto de Telecomunica\u00E7\u00F5es under the projects UIDB/50008/2020 and CELL-LESS6G 2022.08786.PTDC.

Files