Exploring the use of Radio Frequency Characteristics for Industry 4.0 Security

Important results from the project

Minute hardware imperfections in the transmitter causes every device to have a unique radio signature. In this project, we design methods to use these unique radio signatures for device identification. We extend the methods to capture and compare the intrinsic differences between devices. This allows us to identify unseen devices with an accuracy of over 90% using just one radio packet as a reference. This method does not need any modification to incorporate these unseen devices.

Expected long term effects

We developed a contrastive deep-learning method for addressing the network scalability challenges of radio frequency fingerprinting solutions. Our method achieves an accuracy of over 90% in unseen device classification across 44 devices from three different chipsets. We are convinced that our excellent results will also lead to further collaborations with Swedish industry. Moreover, we believe this project will pave the way towards addressing the security challenges in Industry 4.0, thereby enabling the applications for efficient and flexible factory operations.

Approach and implementation

To achieve our objectives, we run a workshop with Swedish industry and technical feasibility studies. Given the very early stage of the project, we were delighted about the interest the workshop gathered. The workshop has highlighted well the challenges and potential of the area. The thorough survey of related work and the technical feasibility studies helped in understanding the key challenges of the technology. This helped us in developing new methods to make the technology more practically applicable.