Demodulation-free protocol identification in heterogeneous wireless networks

Abstract

Nowadays various wireless network protocols play respective roles to fulfill different demands. To better adapt to this heterogeneity and coexistence situation, it is critical for nodes to identify the available networks with high accuracy and low cost. Unlike traditional demodulation-based identification method, which is expensive and complexing, in this paper, we propose a novel conception called demodulation-free protocol identification. This method only employs the features of physical layer samples. We first extract features that can be used to identify different protocols. Specifically, a sparse sequence based Precision-Stable Folding Algorithm (PSFA) is proposed to detect periodicity feature, which is common in wireless network protocols. Then we construct a prototype with USRP to identify three commonly used protocols in the 2.4 GHz ISM band. Experiment results show that under low or moderate channel utilization, the accuracy is above 90%. We also show that the computational complexity is polynomial.

Introduction

The coexistence of heterogeneous networks has become a prominent trend, since various wireless network protocols play respective roles to fulfill different demands. In addition, most of the channels in these networks are overlapping with each other [1]. Take the city shown in Fig. 1 as an example. Wireless Sensor Networks (WSNs) are deployed in hospitals, forests, and roads for data collecting, e.g. CO₂, temperature, pollution, etc., while WiFi hotspots are deployed to provide Internet access in restaurants and campus. In addition, Wireless Personal Area Networks (WPANs) are used for short-distance communications, like smart home networks. In this context, to enhance coexistence and heterogeneity, it is essential for nodes to have a preliminary view of the wireless networks in current region. Therefore, accurate and low-cost protocol identification is playing an important role for quick media access and interoperability.

Traditional protocol identification schemes are demodulation-based. By demodulation and decoding received packets [2], [3], [4], [5], the used protocols can be recognized. This requires nodes to implement all possible network protocol waveforms. The cost is high since physical layer (PHY) and most media access control (MAC) functions are implemented in hardware or firmware. Though Software Defined Radio (SDR) [6] can implement all possible waveforms in software and reduce the cost, nodes still need to load and try each waveform one by one [7]. Besides, packet decoding is not always feasible in practical circumstances, especially under war conditions. Various information technologies (e.g., information encryption) and electromagnetic interference (EMI) will be employed in future high-tech wars. In this situation, the SNR of received signals may drop to a level which cannot satisfy the demodulation requirement.

For above reasons, we are motivated to seek a less expensive protocol identification method, which can use PHY signals only and be demodulation-free. As we know, the current networks are based on a layered architecture, which results in the information scarcity of upper layer protocols when working with only PHY signals. Fortunately, protocol level behavior can be reflected to PHY signals, which leaves us a chance to infer upper-layer protocols through RF analysis. Its advantages are as follows:

• It can reduce the implementation cost. As only PHY signal features are used to recognize different protocols, there is no need to try each demodulation scheme, or implement the whole protocol stack of each potential protocol. This can greatly reduce the implementation complexity and financial cost.

• It can reduce the computational cost. Compared with traditional identification approaches, some signal processing modules are not necessary, such as frequency offset compensation, phase offset compensation, and timing recovery. This reduces the computational complexity.

• It can be used in situations where reliable decoding is not feasible. For example, when scanning with omnidirectional antenna, the received SNR may be low for demodulation. We can first detect the existence of signals without demodulation. Then with beamforming and the direction of arrival estimation, received SNR can be strengthened and interested signals may be able to be demodulated.

Despite of the advantages, it may be more challengeable to consider raw PHY layer samples. Due to the layered architecture of networks, different layers work independently. Thus characterizing different signals and classifying them with these features can be difficult with original PHY layer samples.

Following the above idea, we propose a new conception called demodulation-free protocol identification, which only relies on PHY information. The key contributions of this paper can be summarized as follows:

• We propose the conception of demodulation-free protocol identification. It only employs features of PHY samples. This approach can be embedded into intelligent devices for network identification before media access, and provide interoperability across heterogeneous platforms.

• We investigate and extract the features of PHY signals that can be used to identify different wireless protocols. We analyze different signal features in both time domain and frequency domain. Specifically, a sparse sequence based Precision-Stable Folding Algorithm (PSFA) is proposed to detect the periodicity feature, which is common in wireless protocols [3], [4].

• Taking three commonly used wireless protocols as an example, we construct a system design with USRP [8] to validate the feasibility and performance of the proposed conception. Experiments show that under low or moderate channel utilization ratio, the detection accuracy is above 90% for both single and multiple APs. We also show that the computational complexity is polynomial.

The remainder of this paper is organized as follows. Section 2 presents a review of related works. In Section 3, we investigate the features of different signals in both time domain and frequency domain. Section 4 describes the design and implementation of the identification system. The experimental results are shown in Section 5. Finally, Section 6 concludes the paper and presents future work.