Accurate and Reliable Methods for 5G UAV Jamming Identification With Calibrated Uncertainty

Hamed Farkhari, Joseanne Viana, Pedro Sebastião, Luís Bernardo, Sarang Kahvazadeh, Rui Dinis

Research output: Chapter in Book/Report/Conference proceedingConference contributionpeer-review

Abstract

This research highlights the negative impact of ignoring uncertainty on DNN decision-making and Reliability. Proposed combined preprocessing and post-processing methods enhance DNN accuracy and Reliability in time-series binary classification for 5G UAV security dataset, employing ML algorithms and confidence values. Several metrics are used to evaluate the proposed hybrid algorithms. The study emphasizes the XGB classifier’s unreliability and suggests the proposed methods’ potential superiority over the DNN softmax layer. Furthermore, improved uncertainty calibration based on the Reliability Score metric minimizes the difference between Mean Confidence and Accuracy, enhancing accuracy and Reliability.
Original languageEnglish
Title of host publicationRCIS
Subtitle of host publicationThe 17th International Conference on Research Challenges in Information Science
PublisherCEUR Workshop Proceedings
Publication statusPublished - 2023
EventWorkshop on Research Projects Track at 17th International Conference on Research Challenges in Information, RCIS 2023 - Corfu, Greece
Duration: 23 May 202323 May 2023

Publication series

NameCEUR Workshop Proceedings
PublisherCEUR Workshop Proceedings
Volume3601
ISSN (Print)1613-0073

Conference

ConferenceWorkshop on Research Projects Track at 17th International Conference on Research Challenges in Information, RCIS 2023
Country/TerritoryGreece
CityCorfu
Period23/05/2323/05/23

Keywords

  • 5G
  • 6G
  • Calibration
  • Deep Neural Networks
  • Jamming Identification
  • Reliability
  • Uncertainty
  • Unmanned Aerial Vehicle

Fingerprint

Dive into the research topics of 'Accurate and Reliable Methods for 5G UAV Jamming Identification With Calibrated Uncertainty'. Together they form a unique fingerprint.

Cite this