Enhanced Channel-PUFs: The Advantages of Artificial Intelligence for Channel Estimation

Christoph Lipps, Sachinkumar Bavikatti Mallikarjun, Mathias Strufe, Hans Dieter Schotten

In: 15th International Conference on Cyber Warfare and Security. International Conference on Cyber Warfare and Security (ICCWS-2020) March 12-13 Norfolk Virginia United States Pages 335-345 Academic Conferences and Publishing International Limited Reading 3/2020.


There is a recognizable trend of increasing inter-connectivity, the amount of transmitted data as well as the integration of a multitude of different sensors and actuators into existing (industrial) networks. Together with mobile, flexible and highly scalable amounts of devices, these structures form the Industrial Internet of Things (IIoT), the Internet of Everything (IoE) and Cyber-Physical Production Systems (CPPS). However, the key enabler of this development – wireless communication - comes along with drawbacks in the form of new attack vectors and security threats. Due to the open nature and broadcast characteristic, it is especially prone to miscellaneous cyber-attacks, such as eavesdropping. Just because of this, it is a huge challenge to develop and integrate sound and secure communication between doubtless authenticated entities. Since conventional security depends on complex computations as well as overhead within the communication itself, Physical Layer Security (PhySec) offers a worthwhile solution to face these tasks. Furthermore, the application of Next Generation Mobile Networks (NGMN) such as Long Term Evolution Advanced (LTE+) and the Fifth Generation (5G) are capable to overcome the constraints of bandwidth limitations. 5G with its benefits of high data rates, low latency, an operation in almost real-time and the microcell approach, is a promising solution for future industrial networks. Besides that, the application of Artificial Intelligence (AI) enables new possibilities to utilize inherent given information of the wireless channel and to train models to predict their behaviour and strengthen the generation of a shared secret between two communicating entities. Within this work, PhySec methods are applied to derive symmetric cryptographic credentials and establish a trustful communication. This is done by using a real-world implementation of an NGMN testbed to evaluate the proposed adaptions. AI in the form of the Linear Regression Algorithm (LRA), is applied to enhance the estimated channel model. The results prove the suitability of Channel-PUFs, the benefits of the AI improvements of the Channel Measurement in the Secret Key Generation (SKG) algorithm and the huge overall potential to secure future industrial and private networks. And this, moreover, in a resource efficient and secure manner. The approach is a low cost, efficient and resource-saving alternative to conventional security mechanisms.


German Research Center for Artificial Intelligence
Deutsches Forschungszentrum für Künstliche Intelligenz