Abstract
This paper presents an adaptive, Received Signal Strength (RSS)-based detection mechanism for Primary User Emulation (PUE) attacks in Cognitive Radio Networks (CRNs). By leveraging the significant transmission power disparity between legitimate high-power primary users (e.g., TV towers) and typical low-power attackers, the proposed method utilizes a hypothesis test based on inferred distance consistency. Experimental results from simulations and real-world hardware testbeds demonstrate near-perfect detection rates without requiring GPS or prior training data.
Introduction
Wireless spectrum is a finite and increasingly crowded resource. Cognitive Radio Networks (CRNs) address this scarcity by allowing Secondary Users (SUs) to opportunistically access spectrum when the licensed Primary User (PU)—such as a TV station or radar—is inactive. However, this flexibility introduces a critical security vulnerability: the Primary User Emulation (PUE) attack. In a PUE attack, a malicious actor mimics the signal characteristics of a PU to trick SUs into vacating the band, enabling the attacker to monopolize the spectrum or launch a Denial of Service (DoS) attack.
Existing defenses often rely on precise geographical locations of all users or complex machine learning models requiring extensive "clean" training data. This paper proposes a more robust, adaptive mechanism that relies on physical layer characteristics to distinguish between legitimate PUs and malicious emulators.
Methodology
The proposed detection mechanism is based on Received Signal Strength (RSS) and the "hard-to-mimic" physical feature of transmission power. Legitimate PUs typically transmit at kilowatts (kW), whereas mobile attackers operate in milliwatts (mW). Even if an attacker perfectly mimics a signal's waveform, replicating its sheer magnitude is physically prohibitive.
System Architecture and Grouping
To mitigate the effects of real-world signal fading, the system organizes SUs into local groups based on proximity. A Fusion Center (FC) collects RSS data from these groups and averages them to create a "smoothed" signal profile, as illustrated in the system model above.
Adaptive Hypothesis Testing
The FC performs a hypothesis test ($H_0$ for PU, $H_1$ for attacker) by calculating the "inferred distance" to the transmitter. This calculation assumes the transmitter is a high-power PU.
- Consistency Check: If the transmitter is a real PU, the inferred distances across different SU groups will be geographically consistent.
- Inconsistency Detection: If the transmitter is a low-power attacker, the assumption of high power causes the inferred distances to become wildly inconsistent across the network, triggering an alert.
This approach is inherently adaptive; it supports mobile nodes and does not require SUs to possess GPS or prior knowledge of the PU’s location.
Experimental Evaluation
The authors evaluated the mechanism using both Monte Carlo simulations (100,000 iterations) and a real-world hardware testbed. The hardware setup utilized Universal Software Radio Peripheral (USRP) N210 devices, with a local Digital TV (DTV) station (345 kW) serving as the PU and a USRP device mimicking the DTV signal as the attacker.
Key Results
- Detection Performance: In real-world trials, the system achieved a detection rate near 1.0 with a near-zero false positive rate, primarily due to the 60 dB power differential between the DTV station and the attacker.
- Scalability: Detection accuracy ($P_d$) increases with the number of SUs. With four or more SUs, the detection rate approaches 100% under ideal conditions.
- Comparative Advantage: Compared to a Back-Propagation Neural Network (BPNN), this statistical method remained functional in scenarios where node locations were unknown, a situation where BPNN typically fails.
Discussion and Conclusion
The core innovation of this work is treating physics as a fingerprint. While software identifiers can be spoofed, an attacker cannot easily simulate the signal footprint of a 300-meter tower from a laboratory setting. The "distance inconsistency" logic provides a predictable and detectable metric for identifying imposters.
However, the model's performance is sensitive to the accuracy of the propagation model (e.g., path loss exponent). Future work will focus on refining statistical models to handle extreme signal shadowing in complex urban environments and investigating scenarios where sophisticated attackers might use high-gain antennas to narrow the power gap.
Summary updated