Securing Hybrid Wireless Body Area Networks (HyWBAN): Advancements in Semantic Communications and Jamming Techniques

WBAN is a way to link various wearable sensor nodes wirelessly into one individual and personalized network used to monitor a person’s psycho-physiological vital signs. Depending on the need, the vital sensors can be distributed all around the human body. Low-power consumption, small size, and lightweight are the requirements set for the nodes to enable user acceptance. In principle, the amount of connected sensors within one WBAN can be high, but a realistic number is less than five for the sake of usability. The basic idea behind a WBAN is that dedicated sensors are collecting vital information and transmitting it wirelessly to the central node (called a hub), which then pre-process the data or conveys it further. Figure 2 shows the variability of the vital sensor nodes, which can be used in the WBAN context (the list of sensors is not exhaustive) [13]. In addition to sensors which are attached to the skin, so called on-body sensors, WBAN can utilize smart implants, such as pacemakers, or other in-body sensors/devices, such as Wireless Capsule Endoscope (WCE). In WBAN, all the nodes are connected to the on-body hub to enable real-time information transmission towards backbone infrastructure. Typically, WBAN is using a one-hop star network topology.

Currently, de facto wireless standard in WBAN is Bluetooth Low Energy (BLE) but there are also other dedicated WBAN standards available, such as ETSI SmartBAN [14], IEEE 802.15.6 [15], or IEEE 802.15.4 [16]. The latter one is better known via its higher layer protocols ZigBee and 6LoWPAN.

From a radio technology point-of-view, WBAN connectivity can be based on narrowband (NB) signals, which are used, e.g., in BLE and SmartBAN, or ultra wideband (UWB), which is adopted by [15]. The most common frequency band at the moment for NB signal is Industrial, Scientific and Medical (ISM) band at about 2.4 GHz. On the other hand, e.g., [15] defines several NB frequency bands for WBAN use also occupying sub-GHz frequencies. As operating in a highly populated frequency range, ISM band is typically subject to high interference originated from other radio equipment nearby. The selected frequency band, as well as signal bandwidth, also have an impact on the observed signal propagation properties through/along tissues, positioning accuracy, throughput, etc., depending on the application and use-case. If high resolution, real-time data is needed, then UWB can be the best option from radio-based technologies. Lower performance requirements and deeper in-body penetration, however, are favouring NB technology. Reference [15] also defines Human Body Communications(HBC) technology operating around 21 MHz, but this is omitted in this review due to its deviation from the conventional Radio Frequency (RF)-based communications as being a coupling-based solution.

The original network topology in WBAN is based on a star topology, where all the data flows are going through the central node, the hub. In this case, the hub is also a bridge from the body domain to the backbone network. The recent development, e.g., in ETSI SmartBAN has introduced and defined a hub-to-hub communications to transfer information between adjacent WBAN networks [17]. In addition, a two-hop relay functionality is included in the SmartBAN technical specification [18]. From the security viewpoint, all the hops between WBAN nodes, although being short, should be reliable but also secure as the communications chain is as reliable as its weakest link. This highlights the importance of light security protocols to be used also in the WBAN context.

The utilization of optical communications, including Visible Light Communication (VLC) and Infrared (IR) technologies, in wireless sensor networks is gaining interest for various IoT and body network applications. Optical communication offers security, bandwidth, and energy efficiency advantages, which are crucial for IoT deployments.

VLC utilizes Light Emitting Diodes (LED) to transmit data using the visible spectrum. This approach is inherently secure due to the limited light propagation and offers high data rates, making it suitable for indoor IoT applications [19]. VLC’s potential in hybrid optical-wireless networks for next-generation communications, especially in 5G and beyond, is highlighted in [20]. IR communication leverages the non-visible spectrum for data transmission, offering benefits in terms of device miniaturization and reduced interference with existing RF systems. Its suitability for low-data-rate IoT applications, especially in hybrid networks combining IR and VLC, is explored in [7]. IR in hybrid radio-optical wireless networks offers innovative solutions for versatile IoT applications, as discussed in [21]. Integrating optical and wireless technologies in a hybrid framework opens new avenues for enhancing IoT network performance. The synergy of RF and optical communication technologies in hybrid networks is investigated in [22], which outlines the implementation and advantages of such an approach.

The complexity of HyWBANs, which combines RF and OWC, far exceeds traditional communications systems due to their dual-channel nature. This sophistication poses significant challenges for attackers attempting to compromise the network, as they must navigate radio and optical channels.

In the HyWBANs domain, the security of communication channels is essential, especially when considering transmitting sensitive health data. This section delves into the nuanced vulnerabilities inherent to RF and OWC, laying the groundwork for understanding the superior security posture of optical communications in specific scenarios. RF communications, by their very nature, are susceptible to eavesdropping due to their omnidirectional signal propagation. This characteristic allows malicious entities to intercept signals without necessitating a direct line of sight, thereby posing a significant risk to the confidentiality of transmitted data. Conversely, optical communications demand a line-of-sight for effective transmission, inherently restricting the potential for unauthorized interception. Despite this advantage, optical channels are not impervious to security threats. A breach in the line-of-sight or sophisticated techniques to capture reflected optical signals can compromise data integrity and confidentiality.

In the context of HyWBANs, an attacker would primarily focus on tactics that enable eavesdropping on biomedical device communications. These tactics could include exploiting network security protocol vulnerabilities, conducting Man-in-the-Middle (MitM) attacks to intercept data, or using sophisticated techniques to bypass encryption. Reconnaissance plays a crucial role, as the attacker must gather detailed information about the network’s configuration and security mechanisms to successfully deploy malware or other attack vectors.

Figure 4 presents a multifaceted evaluation of security threats in HyWBANs. Each threat is analyzed based on three critical dimensions: relevance to the network’s security, ease of implementation by potential attackers, and the potential impact on network integrity and functionality. This brief assessment enables a slight understanding of each threat’s effectiveness and helps prioritize security measures.

To fortify the security framework of HyWBANs against these vulnerabilities, we introduce a semantic communication method that significantly enhances the security of transmitted data.

In this study, we investigate the efficacy of hybrid communications that utilize optical and radio channels, explicitly investigating their capacity to penetrate biological tissues. Two distinct experimental setups were designed to assess the performance characteristics necessary for effective and secure communications through such mediums. For optical communications, NIR frequencies were employed, selected for their proven proficiency in penetrating biological tissues. On the other hand, UWB technology, recognized for its superior transmission capabilities, particularly noise-like signals, was chosen for radio communications. This dual-faceted approach allows for a comprehensive evaluation of the potential of hybrid communication systems in medical applications.

The experimental setup, depicted in Fig. 6, is an optical communication part; it comprises various components that can be divided into two subsystems: transmitter and receiver front end. The transmitter unit includes a NIR LED (M810L3, THORLABS) with 810 nm wavelength, a bias-tee, and a LED driver. The receiver unit utilizes a Photo-Detector (PD) (PDA 36A-EC switchable gain detector, THORLABS). A sample of biological tissue was used, acting as the communications channel. The LED is driven by a current driver module (DC2200, THORLABS), which is controlled by an external modulation source. The modulation of the NIR LED is essential for transmitting data through the biological tissue. The PD, positioned at a specific distance (d is the thickness of the meat sample used in the measurements) from the NIR LED, captures the transmitted light after it has passed through the tissue. The output from the PD (\(V_{out}\)) is then analyzed using an oscilloscope (with \(50~\Upomega \) impedance) to assess the effectiveness of data transmission through the tissue sample. Using a laptop, we sent the same ASCII character with the User Datagram Protocol (UDP) protocol to a software-defined radio USRP that modulated the signal before sending it to the LED driver. Two NI USRPs (2920 model) were employed in this study. We also used a bias-tee (ZFBT-4R2GW-FT+, Mini-Circuits) to combine the modulation signal and the bias current to feed the driver. We measured the peak of the received burst signal for each character sent from the laptop. The \(V_{out}\) is then converted into an input power unit by using the equation provided in the datasheet of PDA36A-EC. This setup is crucial for evaluating the feasibility of NIR communications in scenarios where signals need to penetrate biological tissues, such as in implantable medical device applications.

As illustrated in Fig. 7, the radio measurement setup was meticulously designed to evaluate the performance of radio communications in HyWBANs. It consists of a UWB transmitter (P410 PulsON, Time Domain) inside the body (i.e., in-body device) that communicates with a UWB receiver (P410 PulsON, Time Domain) that has its antenna positioned on the porcine skin (i.e., on-body). Using Time Domain’s Channel Analysis Tool (CAT) software, we simulated the communications scenario inside the body by sending signals from the transmitter to the receiver. We enclosed the antennas inside a box of RF absorber material to avoid external interference. The received signals were saved on a laptop using CAT software and analyzed later using MATLAB. This system makes it possible to accurately measure the radio signal’s capability to penetrate biological tissue and the effectiveness of UWB technology in an in-body communications scenario, essential for developing reliable hybrid WBANs.

Figure 8 shows the power (expressed in mW) of the NIR signal received after passing through the two biological tissue samples with a maximum thickness of 37 mm and 39 mm for samples #6 and #7, respectively. From Fig. 8, it is evident how the propagation capabilities improve when the temperature reaches \(37^\circ C\), which is considered almost typical for a human body. Selecting a higher gain (i.e., from 0 dB to 10 dB) in the receiver (PDA 36A-EC offers this option using a rotary switch) does not lead to a significant advantage regarding receiver sensitivity.

Figure 9 shows the UWB SNR measured by the on-body antenna placed on the skin. Meanwhile, the transmitter and receiver were aligned for light. For the UWB measurements, we left the in-body antenna at the fixed position while we moved the on-body antenna in 1 cm steps to investigate the communication limit.

Developing and optimizing semantic communications and strengthening security within hybrid networks necessitate a comprehensive dataset for training and testing DL models. This dataset encompasses both measured features, such as SNR and the power received by the PD in NIR communications, as well as synthetic features. These synthetic features are conceptualized on the premise that the constituent devices of HyWBAN can acquire and process biological signals from individuals. This dual approach in dataset formulation facilitates a realistic assessment of the HyWBAN’s operational capabilities and aids in simulating a wide range of scenarios for advanced medical applications.

To refine the dataset for DL models in the context of HyWBANs for medical applications, we employed a dual-strategy approach involving both the augmentation of measured data and the generation of synthetic data. The augmentation process for the measured attributes, specifically SNR for UWB and received power for NIR, employs a statistical methodology in which new values are generated based on a Gaussian distribution. This distribution is centred on the measured mean and standard deviation. This statistical rigour ensures the augmented data are closely aligned with realistic measurement variations. We generated values for a few parameters: acceleration, heart rate and body temperature to generate synthetic data to emulate the ability of HyWBAN devices to measure biological signals. We assumed that a hybrid device could access these quantities (as a knowledge base for semantic communications) in a medical application or at least a part of it. These synthetic values are derived from a Gaussian distribution, adhering to predefined mean and standard deviations to ensure they fall within physiologically plausible ranges. The Table 1 summarises the specification of statistics for the data generation process. Our semantic communication model uses a binary classification approach, simplifying complex data into categories like ’HIGH_SNR’ or ’LOW_SNR’ using the thresholds defined in Table 1. This method efficiently filters out noise, focusing on key data aspects and significantly reducing computational load. This binary representation accelerates model training and enhances interpretability, facilitating fast and decisive communications analysis. We can then define the labels to be associated with each type of communication in a supervised manner (see Table 2). These labels are the semantic concepts that represent data the device measures in a compressed manner. This particular approach to dataset preparation supports the robustness of the developed models. It ensures the simulation of diverse scenarios, which is critical for applying HyWBAN in medical settings.