Exploring Detection of Unmanned Aerial Systems on 5G Networks Via Machine Learning is a Naval Postgraduate School thesis by Alexander D. Gore.

The advancement and integration of 5G network technologies, encompassing techniques like beamforming and sidelinking, are poised to substantially enhance the capabilities of potential adversaries engaged in drone operations that extend beyond direct line of sight. This study delves into strategies aimed at surmounting the hurdles introduced by 5G concerning detecting drones, especially when data transmission is concealed through encryption.

The methodology involves the creation of datasets about drone network activity. This is achieved through the interception of data packets exchanged between a ground control station and a simulated drone. Subsequently, the distinct communication streams are segregated, and statistical characteristics are formulated utilizing extracted temporal attributes. These attributes encompass metrics such as the mean, median, and standard deviation of inter-arrival times and the ratio of packet directions.

A random forest classifier is trained and evaluated using these derived statistical profiles. This classifier can discern between simulated drone traffic flows propagated through WiFi or Ethernet and regular 5G data flows. The classifier achieves an accuracy rate of 99% and an F1 score surpassing 98% in a fraction of a second. Moreover, the classifier exhibits proficiency in detecting drone-related traffic even when encountered in data transmissions occurring over a distinct system from the one it was initially trained on, maintaining an F1 score exceeding 97%.

While it’s noteworthy that the evaluation did not extend to drone data transmitted over actual 5G networks due to tool limitations, the promising alignment between detection attributes, such as data directional rates, between drone data and typical 5G data remains regardless of the mode of transmission. The proposed approach’s superior performance and its exclusive reliance on temporal attributes establish it as a prospective avenue worth exploring in the realm of 5G drone detection.

Publication Date- June 2023

Exploring Detection of Unmanned Aerial Systems on 5G Networks Via Machine Learning contains the following major sections:

  • Introduction
  • Background
  • Methodology
  • Results
  • Conclusion and Future Work

Approved for public release.  Distribution is unlimited.

C-UAS Hub does not own this content and provides a link for users at the bottom of the page to access it in its original location. This allows the author(s) to track important article metrics related to their work. All credit goes to its rightful owner.

Author- Alexander D. Gore

For additional multimedia resources, please visit the Multimedia Library.

Post Image- Telecommunication tower with 5G cellular network antenna on city background, Global connection and internet network concept (Image Credit: kinwun)