Power Allocation for Fingerprint-Based PHY-Layer Authentication with mmWave UAV Networks

Sung Joon Maeng, Yavuz Yapici, Ismail Güvenç, Huaiyu Dai, Arupjyoti BhuyanICC 2021 · DOI

TL;DR

This paper tackles the challenge of securing mmWave UAV communications against eavesdroppers by proposing a novel power allocation strategy. By jointly optimizing the power distribution for data, artificial noise, and authentication tags, the authors demonstrate a way to balance high data rates with robust authentication and strong secrecy.

What Problem Does This Paper Address?

As unmanned aerial vehicles (UAVs) become increasingly integrated into cellular networks, ensuring secure communication, especially in beyond-line-of-sight (BVLOS) scenarios, is paramount. Traditional security methods like encryption can be computationally intensive and may not be sufficient against sophisticated attacks. Physical Layer Security (PLS) offers a promising alternative by leveraging the inherent characteristics of wireless channels to secure communications. This paper focuses on a specific PLS technique: fingerprint-based authentication, enhanced by artificial noise (AN), within the context of mmWave cellular networks connecting UAVs. The core problem is how to effectively allocate power in such a system to simultaneously achieve high data rates for legitimate users, ensure their authentication, and prevent eavesdroppers from successfully impersonating users or decoding sensitive information. The unique characteristics of mmWave, such as narrow directional beams, offer opportunities for enhanced spatial security, but also present challenges in maintaining reliable connectivity and authentication.

Methods

The authors propose a power allocation strategy that jointly optimizes three key parameters:

  • Power Splitting Factor ($\phi$): This determines how the total transmit power is divided between the data precoder and the artificial noise (AN) precoder.
  • Tag Power Factor ($P_t$): This controls the proportion of power allocated to the authentication tag.
  • AN Precoding Matrix ($\mathbf{V}$): While not directly a power allocation parameter, the design of the AN precoder is crucial for corrupting the eavesdropper's signal. The paper utilizes a null-space precoder for AN, aiming to direct it away from legitimate users.

The authentication mechanism relies on embedding a "fingerprint" (tag) within the transmitted signal. Legitimate users, possessing a shared secret key, can verify their identity by comparing the received tag with their generated expected tag. Eavesdroppers, lacking this key, face a much harder task in both decoding the data and correctly identifying the tag. The paper models the system using a multi-user mmWave scenario with a base station serving multiple UAV users and a single passive eavesdropper. They employ a regularized zero-forcing (RZF) linear precoder for data transmission and analyze the system's performance in terms of authentication probability ($P_D$), correct key detection probability ($P_K$), and sum rate.

Datasets

Not discussed in the paper. The evaluation is based on simulations using parameters provided in Table I.

Key Insights and Ideas

  • Joint Optimization is Key: The paper highlights that optimizing power allocation for data, AN, and the authentication tag jointly is crucial for achieving a desired balance between security and performance, rather than optimizing them in isolation.
  • mmWave's Spatial Advantage: The directional nature of mmWave beams is leveraged to enhance the spatial security of the authentication process, making it harder for eavesdroppers to intercept signals.
  • Trade-offs Exist: There's an inherent trade-off between maximizing the sum rate, ensuring high authentication probability for legitimate users, and minimizing the eavesdropper's ability to detect the correct key (secrecy). The proposed strategy aims to navigate these trade-offs.
  • AN's Dual Role: Artificial noise not only aids in securing the data transmission but also plays a role in corrupting the eavesdropper's ability to accurately detect the authentication tag.
  • Fixed $\omega$ Strategy: The paper introduces a "fixed $\omega$" strategy where a specific relationship between power allocation factors is maintained. This strategy allows for stable authentication and secrecy levels even when improving the sum rate.

Key Results

The paper presents simulation results demonstrating the effectiveness of their proposed power allocation strategies.

  • Authentication Probability ($P_D$): The proposed strategies, particularly the "fixed $\omega$" approach, can achieve high authentication probabilities, often outperforming conventional methods that fix tag power.
  • Correct Key Detection Probability ($P_K$): The proposed strategies allow for control over the eavesdropper's success probability. The "fixed $\omega$" strategy helps maintain a low secrecy level (high $P_K$ convergence) while enabling improvements in sum rate.
  • Sum Rate: The proposed strategies can achieve competitive sum rates, and in some cases, improve them by adjusting power allocation, while maintaining desired security levels. For instance, increasing the power splitting factor ($\phi$) generally increases the sum rate.
  • Relationship between Parameters: The paper visually demonstrates the complex relationships between the power allocation factors ($\psi$, $\omega$, $\phi$, $P_t$) and their impact on $P_D$, $P_K$, and sum rate. For example, increasing $\omega$ (related to secrecy) leads to a decrease in $P_t$ (tag power) and an increase in $\phi$ (data power), impacting the sum rate.

Limitations and Future Work

  • Assumptions: The analysis relies on several assumptions, including the perfect knowledge of the legitimate user's channel and the worst-case scenario where the eavesdropper can decode all users' data.
  • Channel Quality of Eve: The paper notes that the channel quality of Eve is not always known, and the secrecy level can be interpreted as the converged probability of $P_K$.
  • Future Work: The authors suggest that further optimization could involve improving the sum rate at the cost of authentication probability, or vice-versa, depending on specific application requirements. Exploring more complex network topologies and attack models could also be future directions.

Summary updated

BibTeX
@inproceedings{maeng2021power,
  author = {Sung Joon Maeng and Yavuz Yapici and Ismail Güvenç and Huaiyu Dai and Arupjyoti Bhuyan},
  title = {{Power Allocation for Fingerprint-Based PHY-Layer Authentication with mmWave UAV Networks}},
  booktitle = {ICC},
  year = {2021},
  doi = {10.1109/icc42927.2021.9500273},
}
Figure