Approximation-based Threshold Optimization from Single Antenna to Massive SIMO Authentication

Stefan Roth, Aydin Sezgin, Roman Bessel, H. Vincent PoorarXiv 2022 · DOI · arXiv

TL;DR

This paper introduces a Kalman-filter-based authentication protocol for wireless sensor networks (WSNs) that jointly evaluates process data and physical channel integrity. By leveraging Chi-square and Gaussian approximations, the authors develop a method to optimize time-varying thresholds, ensuring a constant security level across diverse hardware setups—from simple single-antenna sensors to massive SIMO (Single-Input Multiple-Output) arrays.

Introduction and Problem Statement

Wireless sensor networks (WSNs) are the backbone of modern industrial monitoring, but they are highly vulnerable to False Data Injection (FDI) attacks. An adversary might infiltrate a sensor or deploy a rogue device to feed a controller "garbage" data, potentially causing physical damage.

System model showing legitimate sensors, malicious sensors, and an adversarial device communicating with a multi-antenna Fusion Center.

The core challenge is the trade-off between security (detecting attackers) and reliability (not rejecting legitimate data). Most existing systems use static thresholds for authentication. However, because wireless channels and physical processes are dynamic, a fixed threshold results in a security level that fluctuates over time. This creates "weak points" that attackers can exploit. The authors set out to create a system that maintains a guaranteed, constant level of security regardless of environmental changes.

Proposed Methodology

The researchers propose a security protocol that operates in three phases: registration, login, and authentication. The architecture integrates process monitoring with channel state information to detect anomalies.

Block diagram of the authentication system, illustrating the dual-track Kalman filtering and joint hypothesis testing process.
  • Dual-Track Kalman Filtering: The fusion center (receiver) runs two Kalman filters simultaneously. One tracks the physical process (e.g., a temperature gradient), and the other tracks the wireless channel state.
  • Joint Hypothesis Testing: When a packet arrives, the system calculates the "innovation"—the difference between the actual measurement and the Kalman filter's prediction. A packet is only authenticated if both the data and the channel measurements fall within specific statistical bounds.
  • Approximation-Based Optimization: To handle different hardware scales, the authors provide three mathematical frameworks: Chi-Square Approximation (low-dimensional), Gaussian Approximation (high-dimensional/massive SIMO), and Asymptotic/Channel Hardening (deterministic channel scenarios).
  • Iterative Thresholding: Instead of a fixed number, the system uses these approximations to iteratively calculate a unique threshold ($\eta$) for every single packet to ensure the probability of an adversary succeeding stays exactly at a target level (e.g., 5%).

Key Insights

  • The "Weakest Link" Principle: The authors argue that since adversaries look for the easiest moment to strike, a system is only as secure as its most vulnerable time instant. This justifies the need for time-varying thresholds.
  • Joint Integrity is Key: By checking both the content (process data) and the context (channel state), the system creates a "double-lock." An external attacker might spoof the data but will have the wrong channel; an internal attacker has the right channel but will likely inject anomalous data.
  • Scalability via Math: The transition from Chi-square to Gaussian approximations allows the same security logic to scale from a simple IoT thermometer to a high-end 5G base station with hundreds of antennas.

Experimental Results

The paper relies on synthetic simulations to validate the mathematical models, simulating a Gauss-Markov process system with a $10 \times 10$ state matrix and varying antenna counts (up to $M_C = 100$).

  • Constant Security Achieved: Numerical results show that while a constant threshold leads to a "rollercoaster" of security levels, the proposed time-varying thresholds successfully maintain a constant security level over time.
  • Approximation Accuracy: The Chi-square approximation was found to be highly accurate for small antenna arrays, while the Gaussian approximation converged to the true performance as the number of antennas increased.
  • Reliability Trade-off: The study demonstrated that to keep security constant, the system must occasionally accept lower reliability (higher false rejection of legitimate packets) during periods of high environmental uncertainty.

Conclusion and Future Work

While effective, the approach assumes the process and channel follow Gauss-Markov models. If the underlying dynamics are non-linear, standard Kalman filters might underperform. Additionally, calculating optimized thresholds iteratively adds a computational burden to the fusion center. Future work could explore "game-theoretic" scenarios where smart adversaries attempt to exploit the receiver's thresholding logic.

Summary updated

BibTeX
@misc{roth2022approximation,
  author = {Stefan Roth and Aydin Sezgin and Roman Bessel and H. Vincent Poor},
  title = {{Approximation-based Threshold Optimization from Single Antenna to Massive SIMO Authentication}},
  year = {2022},
  eprint = {2208.05991},
  archivePrefix = {arXiv},
  url = {https://arxiv.org/abs/2208.05991},
}
Figure