PSAU-Defender: A Lightweight and Low-Cost Comprehensive Framework for BeiDou Spoofing Mitigation in Vehicular Networks

Abstract

The increasing reliance of Vehicular Ad-hoc Networks (VANETs) on the BeiDou Navigation Satellite System (BDS) for precise positioning and timing information has raised significant concerns regarding their vulnerability to spoofing attacks. This research proposes a novel approach to mitigate BeiDou spoofing attacks in VANETs by leveraging a hybrid machine learning model that combines XGBoost and Random Forest with a Kalman Filter for real-time anomaly detection in BeiDou signals. It also introduces a geospatial message authentication mechanism to enhance vehicle-to-vehicle (V2V) and vehicle-to-infrastructure (V2I) communication security. The research investigates low-cost and accessible countermeasures against spoofing attacks using COTS receivers and open-source SDRs. Spoofing attack scenarios are implemented in both software and hardware domains using an open-source BeiDou signal simulator to examine the effects of different spoofing attacks on victim receivers and identify detection methods for each type, focusing on pre-correlation techniques with power-related metrics and signal quality monitoring using correlator values. The emulation results demonstrate the effectiveness of the proposed approach in detecting and mitigating BeiDou spoofing attacks in VANETs, ensuring the integrity and reliability of safety-critical information. This research contributes to the development of robust security mechanisms for VANETs and has practical implications for enhancing the resilience of transportation systems against spoofing threats. Future research will focus on extending the proposed approach to other GNSS constellations and exploring the integration of additional security measures to further strengthen VANET security.

1. Introduction

VANETs are emerging as a transformative technology with the potential to revolutionize transportation systems by enabling vehicles to communicate with each other (vehicle-to-vehicle) and with roadside infrastructure (vehicle-to-infrastructure) [1]. This interconnectedness promises enhanced road safety through real-time hazard warnings, optimized traffic flow for reduced congestion, and improved fuel efficiency. Nonetheless, the reliance of VANETs on wireless communication channels exposes them to a myriad of security threats, including spoofing attacks, which can have catastrophic consequences due to the safety-critical nature of the information exchanged within these networks. Thus, the security of VANETs is not merely a technological concern; it is a matter of paramount importance due to the potential impact on human lives and the integrity of safety-critical information. A successful spoofing attack can manipulate a vehicle’s perceived location, velocity, or time, leading to disastrous consequences such as phantom collisions, erroneous navigation instructions, and compromised traffic control systems [2]. To mitigate these risks, robust security mechanisms must be integrated into VANETs to ensure the confidentiality, integrity, and availability of transmitted data.

The BDS has become the basis of modern navigation to offer pinpoint accuracy and timing information which is vital to various fields and industries. Its significance is particularly pronounced in regions like Al-Kharj (Saudi Arabia) and Lahore (Pakistan), where it underpins critical applications in transportation, logistics, and agriculture. BeiDou’s triple-frequency signals and sophisticated capabilities have garnered widespread adoption [3]. Yet, a critical vulnerability lies in the fact that the core BeiDou signals utilized for navigation are not encrypted [4]. The absence of encryption in BeiDou’s core navigation signals exposes them to a range of adversarial manipulations. Malicious actors can exploit this vulnerability by crafting counterfeit signals that closely resemble legitimate BeiDou signals. These deceptive signals can mislead receivers into computing erroneous positions, velocities, or timing information, potentially causing havoc in VANETs. For instance, a spoofed BeiDou signal could misrepresent a vehicle’s location (i.e., as exhibited in Figure 1), leading to incorrect collision warnings or facilitating malicious route guidance.

The efficacy of spoofing attacks is contingent upon several factors, including the computational capabilities of V2I nodes, real-time response imperatives, and the quality & availability of data [5]. Nodes are often constrained by limited processing power and may encounter challenges in executing computationally demanding anti-spoofing algorithms in real time. The time-sensitive nature of V2I communication necessitates swift detection and response mechanisms to thwart spoofing attacks and avert immediate safety hazards. Likewise, the performance of spoofing detection algorithms is inherently linked to the consistency and reliability of the data received from the network’s diverse array of sensors. Variabilities in data quality or periods of unavailability can hinder the algorithms’ ability to accurately discern legitimate signals from fraudulent ones, which can potentially compromise the overall security of the network. To comprehend the impact of BeiDou spoofing attacks, we established a mathematical framework with a vision that it would quantify the impact of spoofing attacks, model relationships between factors, predict consequences, and optimize defenses that eventually will lead us to provide a structured approach for understanding and mitigating risks. Consequently, the received BeiDou signal at a VANET node is denoted as follows:

$$s\left(t\right)={\sum }_{k=1}^N\sqrt{2P_c}{h}_k{D}_k\left(t-{\tau }_k\right)x_k\left(t-{\tau }_k\right)\mathrm{c}\mathrm{o}\mathrm{s}(2\pi \left(f_c+f_{d,k}\right)t+\delta {\theta }_k)$$

(1)

where $P_c$ is the received signal power, $h_k$ represents the channel matrix for the number of satellites, $D_k\left(t-{\tau }_k\right)$ is the navigation data, $x_k\left(t-{\tau }_k\right)$ denotes the spread spectrum sequence, $f_c$ is the carrier frequency, $f_{d,k}$ represents the Doppler frequency shift for each satellite, and $\delta {\theta }_k$ is the carrier phase offset. Our in-depth investigation revealed that the phase angle differences [6] can also play a crucial role as they can indicate discrepancies between authentic and spoofed signals. The phase angle difference between the authentic BeiDou signal and the spoofed signal is expressed as follows:

$$∆\theta =2\pi \left(f_c+f_{d,s}\right)t+\delta {\theta }_s-(2\pi \left(f_c+f_{d,a}\right)t+\delta {\theta }_a$$

(2)

where $f_{d,s}$ and $\delta {\theta }_s$ are the Doppler frequency shift and carrier phase offset of the spoofed signal, respectively, and $f_{d,a}$ and $\delta {\theta }_a$ are the corresponding values for the authentic signal.

This study offers several significant advancements in the field of security for BeiDou-enabled VANETs. Firstly, it introduced an innovative technique to counter spoofing attacks by integrating a hybrid machine learning framework that combines XGBoost and Random Forest algorithms with a Kalman Filter, facilitating real-time anomaly detection in BeiDou signals. More importantly, it incorporated a geospatial attribute-based message authentication mechanism which was optimally tailored to boost the security of V2V and V2I communication. The paper further explores cost-effective and accessible anti-spoofing strategies employing commercially available off-the-shelf (COTS) receivers and open-source software-defined radios (SDRs) (herewith, the SDRs are communication systems where traditional hardware components such as mixers, filters, amplifiers, modulators/demodulators, and detectors are implemented through software on either a personal computer or an embedded system. This adaptability enables SDRs to be reprogrammed or adjusted to function across diverse frequencies and protocols, enhancing their utility for experimental research. They are especially valuable in settings that involve the testing and simulation of spoofing attacks within vehicular networks, including those that use BeiDou signals). The research implemented spoofing attack scenarios in practical settings by leveraging an open-source BeiDou signal simulator in both software and hardware contexts. This method facilitated a detailed analysis of different spoofing attacks’ effects on victim receivers, enabling the identification of tailored detection techniques for each scenario. It specifically emphasized pre-correlation techniques that utilize power-related metrics and evaluate signal quality through correlator values.

The research paper progresses with a ‘Literature Review’ that examines existing studies on BDS spoofing and VANET security by identifying key shortcomings and underscores the necessity for effective anti-spoofing strategies. In Section 3, a cutting-edge approach is outlined, combining a hybrid machine learning model—featuring XGBoost, Random Forest, and a Kalman Filter—with a sophisticated cryptographic method, Attribute-Based Encryption. The ‘Emulation Setup and Assessment Outcome’ thoroughly tests this framework, leading to a ‘Conclusion’ that highlights the study’s contributions and suggests future research avenues to enhance security measures.

2. Literature Review

It is evident that the VANETs are pivotal for the advancement of intelligent transportation systems (ITS) by offering tremendous potential to enhance road safety and traffic efficiency. The integration of accurate positioning systems such as GPS and BeiDou plays a crucial role in enabling functionalities like automated tolling, dynamic routing and collision avoidance in these networks. The inclusion of China’s BeiDou satellite system into the global navigation satellite systems (GNSS) array offers diversification in positioning sources which is crucial for the resilience and reliability of ITS applications. Nonetheless, the increasing reliance on GNSS also introduces significant security challenges, notably in the form of spoofing attacks which could undermine the operational integrity of VANETs.

As exhibited in

Table 1. BeiDou Navigation Satellite System constellation specifications for VANETs with considerations for BeiDou spoofing [7].

Feature/Parameter	BeiDou GEO	BeiDou IGSO	BeiDou MEO	Supplementary Notes (VANETs & Spoofing)
Altitude (km)	~35,786	~35,786	~21,528
Located (longitude)	80° E, etc.	110.5° E, etc.	Varied	Precise locations vary for MEO.
Satellite Clock	High precision	High precision	High precision	Atomic clocks. Vulnerable to spoofing; authentication methods are crucial for VANETs.
NAV Messages	B1I, B2I, etc.	B1I, B2I, etc.	B1I, B2I, etc.	Contain ephemeris, clock data, etc. Integrity monitoring essential to detect spoofed messages.
Frequency (MHz)	B1I (1561.098)	B1I (1561.098)	B1I (1561.098)	B2I and B3I are also available. Signal authentication is key for VANETs to resist spoofing.
Signal Types	B1I, B2I, etc.	B1I, B2I, etc.	B1I, B2I, etc.	Different modulation schemes.
Positioning				High accuracy important for VANET safety.
Navigation				Vulnerable to spoofed signals leading to incorrect directions.
Timing				Precise timing crucial for VANET coordination.
Message Types (Decimal)	0–10			Include ephemeris, clock corrections, etc.
	0: Reserved 1: BGD (Broadcast Navigation Message 2: SBAS Data (System Service Message) 3: IFM (Ionospheric Delay Model) 4: Tropospheric Delay Model 5: UTC(n)—UTC(n-1) 6: Almanac 7: User Range Accuracy Index 8: Satellite Clock Bias 9: Satellite Differential Clock Bias 10: Satellite Orbit Correction
SIS Status	Healthy; Unhealthy; Marginal; Unknown; Don’t Use.			Monitoring signal health to detect anomalies indicative of spoofing.
Signal Integrity Flag (SIF)				Indicates signal quality. 0: Signal is healthy 1: Signal is unhealthy or unreliable
Data Integrity Flag (DIF)				Indicates data reliability. 0: Data OK 1: Data Suspect

, BeiDou systems face increased risks of spoofing attacks due to several factors including their unique signal structures and the strategic importance of the regions they serve. BeiDou operates with a complex signal architecture which presents specific challenges in terms of signal authentication and integrity verification. The dependency of certain regional infrastructures on BeiDou exacerbates these challenges, making it a significant target for spoofing activities. Besides, compared to GPS, BeiDou systems often lack extensive commercial solutions for spoofing detection which complicates the implementation of robust countermeasures.

Murad et al. (2024) [8] introduced a novel GPS-IDS framework designed to detect spoofing attacks on autonomous vehicles (AVs). The framework leveraged a unique physics-based vehicle behavior model that integrated a navigation model to accurately capture the lateral dynamics and states of an AV. By extracting temporal features from examined prototypes and employing machine learning models for classification, the framework achieved an F1 score of up to 94.4%, with a detection time improvement of up to 13 s compared to the Extended Kalman Filter (EKF) detector. To facilitate projected research, a new dataset, AV-GPS-Dataset, was created, capturing real-world GPS spoofing attacks on an autonomous vehicle testbed. This dataset, along with the physics-based model, represented a significant contribution to the field. Nonetheless, the performance of the machine learning models when trained on smaller datasets was a limitation, indicating potential challenges in generalizing to unseen data.

Bethi et al. (2021) [9] proposed a novel method to mitigate the effects of GPS spoofing attacks on navigation systems. The authors developed a robust positioning algorithm that accepts both authentic and spoofed GPS signals, utilizing an iterative least squares (ILS) approach to estimate the receiver’s position with all combinations of measurements. To reduce the algorithm’s complexity, they introduced an M-best position algorithm that selects the M most likely positions at each epoch based on a likelihood-based cost function. The authors then integrated a Kalman filter (KF) with a gating technique to estimate the vehicle’s time-varying dynamics and eliminate false position estimates resulting from incorrect measurement associations in the ILS. The KF framework incorporated nearest neighbor (NN) and probabilistic data association (PDA) algorithms to enhance the association of measurements to the correct track. The proposed algorithm’s performance was evaluated through simulations in both open-sky and urban environments, with varying numbers of authentic and spoofed signals. The results demonstrated the algorithm’s effectiveness in mitigating spoofing effects, particularly in scenarios with higher satellite visibility. The authors also acknowledged the limitations of their approach in handling non-ideal spoofer scenarios and non-Gaussian measurement noise, suggesting these as areas for future research. A notable strength of this research lies in its innovative approach to mitigating GPS spoofing by incorporating both robust positioning and KF-based tracking with data association techniques. The M-best position algorithm effectively reduces the computational complexity of the robust positioning process, making it more feasible for real-time applications. The use of a gating technique within the KF framework further enhances the algorithm’s ability to reject false position estimates. Yet, the study’s primary limitation is its assumption of an ideal spoofer, which may not hold in real-world scenarios where spoofers can employ more sophisticated techniques. Moreover, the algorithm’s performance in urban environments with multipath signals was shown to be less effective, highlighting the need for further research in this area.

Chen et al. (2020) [10] introduced a new methodology for GNSS spoofing detection that leverages two low-cost antennas and eliminates the need for an Inertial Measurement Unit (IMU). The authors developed a hypothesis test based on the sum of squared errors (SSE) derived from carrier phase double difference data and broadcast ephemeris information. The efficacy of this method was rigorously assessed in both static and dynamic scenarios, showcasing its ability to discern spoofing signals originating from single or multiple sources and diverse directions. A notable strength of this research lies in its cost-effectiveness and simplicity, as it circumvents the need for expensive IMUs and intricate attitude determination solutions. The method’s capacity to detect spoofing signals from multiple directions marks a significant advancement compared to traditional methods that often operate under the assumption of a single spoofing source. Yet, the study’s limitations include potential instability in dynamic scenarios due to signal lock-lose and a reliance on prior knowledge of the baseline length, a condition that might not always be met in real-world applications. Likewise, the evaluation primarily focused on meaconing attacks, leaving the method’s performance against more sophisticated spoofing techniques unexplored.

The research presented by David et al. (2022) [11] introduced a sophisticated GNSS spoofing detection method designed for electrical substations (i.e., not directly related to VANETs but furnished in-depth knowledge for our investigation), leveraging commercial off-the-shelf equipment to ensure cost-effectiveness and immediate deployment capabilities. The method’s core innovation lies in utilizing multiple GNSS receivers placed near each other within substations, making it difficult for attackers to send unique spoofing signals to each antenna. In an attack scenario, this setup would cause multiple antennas to report improbable locations, alerting operators to switch to backup timing sources. Notably, this method’s strength is its immediate applicability to existing substation infrastructure, requiring no specialized hardware and minimizing operational disruptions during implementation. Nonetheless, limitations exist. The effectiveness of the method relies on the physical arrangement of antennas and the assumption that spoofing attacks lack the sophistication to target individual antennas subtly. Moreover, the reliance on close antenna proximity may not be optimal for large substations where a wider distribution might be advantageous. These factors suggest that although the method is effective in specific scenarios, it requires further refinement to counteract sophisticated spoofing techniques that can circumvent the existing setup.

A novel Global Positioning System spoofing detection technique that leverages crowd-sourced information from mobile cellular infrastructure and WiFi networks was proposed and investigated by Oligeri et al. (2022) [12]. The authors argued that current solutions for GPS spoofing detection either rely on accessing the physical properties of received GPS signals, which is often not possible with commercial GPS receivers, or on cross-checking with information from a single additional source, which can be unreliable. The approach involved comparing the information received from the GPS infrastructure with that from nearby mobile cellular base stations and WiFi access points, whose positions are publicly available through platforms like OpenCellID and WiGLE. The authors conducted an extensive measurement campaign, driving around for 5 h and covering over 196 km, to gather data from real GSM infrastructure and local WiFi networks. This data was used to build analytic models for the number of in-range base stations and access points, the distance between the user and the anchors, the received signal strength, and the location estimation error. These models were then used to evaluate the performance of the projected technique against GPS spoofing attacks. The authors’ solution allowed for a tunable trade-off between detection delay and false positives. For example, they could detect an attack in approximately 6 s using WiFi information alone, while the delay increased to 30 s when using information from the mobile cellular network, still maintaining a false-positive probability of less than 0.01. The authors also examined the constraints and limitations of their method for the minimum level of accuracy in detection, time requirements, and resistance to adverse data. The study’s strengths include its innovative use of crowd-sourced information from multiple sources, extensive experimental dataset, and the ability to tune the trade-off between detection delay and false positives. Yet, the limitations include the reliance on the availability and accuracy of crowd-sourced data, the potential for malicious information to be injected into these sources, and the varying performance depending on the density of cellular base stations and WiFi access points in different areas.

In another novel study, Vitale et al. (2021) [13] proposed a secure architecture to enhance end-to-end verification of transmitted data among entities in connected and autonomous vehicles (CAV) scenarios. The architecture encompassed a Public Key Infrastructure (PKI) for certificate distribution and updates, a multi-RAT communication infrastructure with Multi-access Edge Computing (MEC) functionalities for computation capabilities and cross-technology communication (e.g., 802.11p and LTE), and a tamper-resistant On-Board Unit (OBU) with an anti-hacking device for running machine learning algorithms. The research focused on detecting and mitigating GPS location spoofing attacks. The authors proposed two approaches for attack detection: an in-vehicle method using Bayesian filtering and a collaborative method leveraging the CARAMEL infrastructure. The in-vehicle method created a fallback localization technique using on-board sensors and Signals of Opportunity (SoO), comparing it with GPS measurements to detect discrepancies. The collaborative method involved vehicles sharing measurements with an ITS application in the MEC, which then employed robust localization algorithms to identify inconsistencies. The study demonstrated the effectiveness of these approaches in simulations using the CARLA simulator, achieving high detection rates for GPS spoofing attacks. However, the research was limited to simulations and did not include real-world testing. Also, the projected countermeasure, revoking certificates of attacked vehicles, required further investigation regarding its overhead on network traffic and scalability with many compromised vehicles.

Table 2. Comparative Analysis of Existing Spoofing Mitigation Techniques in GNSS-based Systems.

Criteria	Murad et al. (2024) [8]	Bethi et al. (2021) [9]	Chen et al. (2020) [10]	David et al. (2022) [11]	Oligeri et al. (2022) [12]	Vitale et al. (2021) [13]
Targeted GNSS	GPS	GPS	GNSS	GNSS	GPS	GPS
Data Sources	GPS data, vehicle dynamics	GPS signals, iterative calculations	Two-antenna system	Multiple GNSS receivers	Mobile cellular and WiFi signals	On-board sensors, external signals
Detection Mechanism	Physics-based vehicle behavior model, machine learning	Robust positioning algorithm, iterative least squares	Sum of squared errors from carrier phase double difference data	Commercial equipment, physical layout of receivers	Crowd-sourced information analysis	Bayesian filtering, collaborative measurements
Mitigation Strategy	Anomaly detection, behavior modeling	Position estimation with gating technique	Spoofing signal discrimination	Improbable location detection, backup timing sources	Comparative data analysis	Secure architecture, data verification
Evaluation Metrics	F1 score, detection time improvement	Performance in open-sky and urban environments	Ability to detect from multiple directions	Applicability to existing infrastructure	Detection delay vs. false positives	Detection rates, simulation effectiveness
Real-world Applicability	Limited by smaller dataset sizes	Simulations in varying environments	Tested in static and dynamic scenarios	Immediate applicability to substation infrastructure	Extensive measurement campaign	Simulations using CARLA simulator
Scalability	Challenges in generalizing to unseen data	Limited by computational complexity of algorithm	Does not require an IMU, simpler setup	Scalable within substation sizes	Depends on the density of cellular and WiFi networks	Requires extensive infrastructure for certificate management
Technique	Machine learning classification	M-best position algorithm with data association techniques	Hypothesis testing based on antenna data	Multiple receivers in close proximity	Comparison of GPS and local network data	PKI, multi-RAT communication infrastructure
Key Features	Integration of vehicle dynamics with GPS data	Gating technique to eliminate false positions	Low-cost, uses simple equipment	Utilizes redundancy in physical equipment setup	Utilizes public location data for verification	Combines on-board and network data for localization
Advantages	High detection accuracy with appropriate dataset	Effective in diverse signal environments	Cost-effective and easy to deploy	Minimal disruption during implementation	Adaptable to various real-world conditions	High security through end-to-end verification
Limitations	Performance drop with smaller datasets	Assumption of ideal spoofer	Unstable in highly dynamic environments	Requires specific antenna arrangement	Accuracy dependent on external data quality	Complex setup, high network load

reiterates a detailed comparison of various spoofing mitigation techniques employed in GNSS-based systems as proposed by different researchers over the years. It highlights the unique strategies, advantages, and limitations of each method, offering insights into how they address the challenges of spoofing in vehicle navigation and control systems.

During our literature investigation, we recognized that a deep understanding of GNSS and specifically GPS spoofing patterns was technically crucial for developing a novel BDS spoofing mitigation model. This knowledge allowed us to dissect the intricacies of spoofing attacks (e.g., as illustrated in

Table 3. Technical weaknesses in GNSS facilitating navigational spoofing attacks in VANETs.

Parameter	Hardware Affected	Software Affected	Protocol/Signal Affected	Impact on VANETs
Signal Strength Fluctuation	GNSS Receivers	Spoofing Detection Algorithms	GNSS Signal Integrity Checks	Increases susceptibility to spoofing by mimicking signal strength variations.
Time Synchronization Errors	Timing Modules	Time Synchronization Software	NTP, PTP	Causes delays in real-time communication, affecting safety-critical applications.
Ephemeris Data Corruption	Navigation Message Processors	Navigation Software	GNSS Data Protocols	Leads to incorrect positioning information, disrupting route and speed calculations.
Non-Encrypted Signals	GNSS Antennas, Receivers	Security Modules	GPS L1, GLONASS L1, BeiDou B1, etc.	Allows easy manipulation and replay of authentic GNSS signals without detection.
Spoofing Signal Injection	RF Front Ends	Anti-Spoofing Filters	Spread Spectrum Techniques	Facilitates the introduction of false signals, misleading navigation systems.
Antenna Placement and Design	GNSS Antennas	Antenna Management Software	Signal Reception Quality	Poor placement and/or design can exacerbate vulnerability to spoofed signals.
Phase Noise	Oscillators, Frequency Synthesizers	Signal Processing Software	Carrier Frequency Stability	Affects the reliability of signal detection under spoofing attack.
Carrier Phase Errors	Frequency Synthesizers	Signal Alignment Software	Carrier Phase Measurement	Misalignments can mislead precise positioning, critical for accurate navigation.
Code Tracking Errors	Signal Processors	Tracking Algorithms	Code Lock Indicators	Disrupts the receiver’s ability to maintain accurate tracking of GNSS signals.
Multipath Interference	GNSS Antennas, Receivers	Multipath Mitigation Software	Signal Propagation Paths	Reflective surfaces create erroneous readings, potentially exploited by spoofers.

), revealed their underlying mechanisms, signal characteristics, and vulnerabilities. By analyzing the unique patterns and behaviors exhibited by GNSS spoofing signals, we could identify potential weaknesses that can be exploited in the design of effective countermeasures. This insight enabled us to develop a sophisticated algorithm and technique that can accurately detect and mitigate BDS spoofing attempts, ensuring the integrity and reliability of navigation and positioning systems that rely on BeiDou signals.

3. Proposed Methodology

We initiated our research with a basic theorem, “In addressing the challenges posed by spoofing attacks within BeiDou-enabled VANET, what methodologies and technological innovations can be employed to not only detect and identify such security breaches effectively but also enhance the system’s tolerance and management capabilities against these threats?” This question was of critical importance as the widespread adoption of BeiDou in VANETs necessitates robust security measures to ensure the safety and reliability of these networks. The unique characteristics of BeiDou signals, such as their triple-frequency structure and regional adoption patterns, present distinct challenges in spoofing detection and mitigation, compared to other GNSS constellations. The triple-frequency nature of BeiDou signals, while offering potential advantages in terms of accuracy and robustness, also introduces complexities in signal processing and spoofing detection algorithms.

3.1. Data Sources and Sampling Strategy

We employed three primary data types to effectively address spoofing in V2I: real-time navigational data from BeiDou systems, network traffic logs, and security incident reports. These data types were carefully chosen for their direct relevance to the operational dynamics of VANETs and their ability to expose patterns in spoofing activities. For example:

(a)

Real-time navigational data was crucial for observing how spoofing could alter positioning information.

$$S\left(t\right)=N\left(t\right)+\delta \left(t\right)$$

(3)

where $S\left(t\right)$ represents the spoofed signal received, $N\left(t\right)$ is the normal BeiDou signal, and $\delta \left(t\right)$ represents the spoofing distortion at time $t$. Equation (3) highlights the deviation caused by spoofing, which can be detected by analyzing discrepancies from expected BeiDou navigational data.

(b)

Network logs provided insights into the frequency and types of network interactions which were essential for detecting anomalies that might indicate spoofing activities.

$$A\left(f\right)={\sum }_k^K{L}_k. e^{-i2\pi fk/T}$$

(4)

where A(f) represents the amplitude of the frequency component $f$ in the network traffic, $L_k$ are the log entries, and $T$ is the total logging period. Equation (4) highlights the Fourier transformation that helped in identifying unusual frequency components in network traffic, suggesting potential spoofing activities.

(c)

Security incident reports offered historical context that aided in understanding past attacks and predicting future vulnerabilities.

$$R={\displaystyle \frac{1}{M}}{\sum }_{m=1}^M{I}_m$$

(5)

where $R$ calculates the risk score based on past security incidents, $I_m$ is the impact score of the $m$-th incident, and $M$ is the total number of incidents reported. Equation (5) represents the aggregation of historical security data to assess the risk and pattern of spoofing attacks, aiding in the prediction of vulnerabilities.

3.2. Data Analysis and Preparation

The proposed framework (i.e., as illustrated in Figure 2) employed data collection using advanced tools such as BeiDou signal receivers (i.e., Trimble BD982 GNSS Receiver, Septentrio AsteRx4, NavCom SF-3050, and u-blox ZED-F9P) and sophisticated network monitoring software (i.e., PRTG Network Monitor 24). These instruments were crucial for capturing real-time navigational data and network traffic, which are essential for analyzing spoofing activities. The sampling strategy involved diverse vehicular types (i.e., passenger cars, commercial trucks, and busses) and network environments (i.e., urban/highways/rural/sub-urban/intersections & traffic lights/covered roads/parking lots/toll plaza environments) to ensure the data encompass a wide range of potential spoofing scenarios (e.g., signal replay attacks, meaconing attacks, false signal generation, signal overlay attacks, record and playback attacks, time synchronization spoofing, ephemeris data spoofing, signal power manipulation, carrier phase spoofing, code delay spoofing, Sybil attacks in network layers). The sample size was precisely chosen based on statistical power analysis (i.e., priori power analysis; post-hoc power analysis; compromise power analysis; sensitivity analysis; and conditional power analysis) to guarantee that the dataset is sufficiently large to validate the research hypotheses while being representative of varied network scenarios and attack vectors. For the statistical analysis of data representativeness, we applied the following model to determine the minimum sample size required for reliable detection capabilities:

$$n\ge ({\displaystyle \frac{Z_{\alpha /2^{\sigma }}}{E}}{)}^2$$

(6)

In Equation (6), n denotes the sample size, $Z_{\alpha /2}$ is the critical value from the standard normal distribution for a confidence level $\alpha , \sigma$ represents the standard deviation of the observed data, and $E$ is the margin of error. This modeling ensures that the selected sample size is statistically valid for conducting effective spoofing detection within diverse network conditions.

3.3. Applied Analytical Technique for Spoofing Detection

To ensure comprehensive data analysis, both quantitative and qualitative methods were employed. For quantitative data, advanced signal strength indicators (i.e., Carrier-to-Noise Ratio (C/N0) Meter; Automatic Gain Control (AGC); Received Signal Strength Indicator (RSSI); Signal Quality Monitoring; and Spectrum Analyzer Output) and a precise time-stamp logging technique (i.e., Precision Time Protocol (PTP)) were utilized to measure and record the navigational data accurately. This allowed for the monitoring of real-time changes in signal properties potentially indicative of spoofing activities.

$$Signal Integrity=\sqrt{{\sum }_{i=1}^n(S_i-\overline{S}{)}^2}$$

(7)

In Equation (7), $S_i$ represents individual signal measurements and $\overline{S}$ is the average signal strength.

For qualitative analysis, we developed detailed procedures to document discrepancies in navigation reports (e.g., inaccurate distance calculations; time discrepancies; incorrect speed readings; anomalous route suggestions; sudden loss of BeiDou signals; unexpected location jumps; inconsistent altitude data, etc.) and feedback from vehicle drivers. This involved setting up protocols for participants’ observation and data collection (i.e., real-time monitoring of participant behavior; use of onboard cameras for visual data recording; tracking for movement & route verification; and use of mobile application for capturing real-time feedback), where individuals involved in the transportation network were observed under normal and simulated spoofing scenarios to capture their responses and system behaviors effectively. These practices facilitated a dual-layered approach to data capture and analysis, which enabled a robust examination of the potential impacts and patterns.

It is worth describing that the data preparation was a crucial phase which involved accurate cleaning and normalization processes (i.e., removal of outliers; filling missing values; standardization of data scales; and conversion of categorical data to numerical format) to ensure the accuracy and reliability of the dataset used for analysis. We employed Python’s Pandas libraries for handling and processing large datasets which allowed for effective manipulation and preparation of data. For quantitative analysis, statistical software MATLAB (version: R2024a), alongside machine learning platform TensorFlow was utilized to apply statistical tests and develop the predictive model.

$$X_{norm}={\displaystyle \frac{X-\mu }{\sigma }}$$

(8)

In Equation (8), $X_{norm}$ denotes the normalized data, $X$ is the original data, $\mu$ is the mean, and $\sigma$ is the standard deviation, facilitating standardization across the dataset. Likewise, data privacy was ensured through encryption and secure data handling practices, compliant with international standards (i.e., General Data Protection Regulation (GDPR)). All sensitive information was anonymized (i.e., by applying techniques such as pseudonymization of identifiers; data shuffling; differential privacy; data masking; generalization of data points; and encryption of data fields) before analysis to protect the privacy of individuals involved in the study.

To effectively enhance anomaly detection, we implemented a hybrid machine learning model by specifically incorporating the XGBoost and Random Forest algorithms integrated with a Kalman Filter. The rationale behind this selection was rooted in the robust capability of these algorithms to manage large datasets and identify subtle, complex patterns in real-time navigational data, which is crucial for spotting sophisticated spoofing attempts. Herewith, XGBoost applied a gradient boosting framework that built decision trees sequentially, where each subsequent tree corrected errors made by the previous ones. This method was particularly effective in enhancing predictive accuracy, as described in Equation (9):

$$F_{t+1}\left(x\right)=F_t\left(x\right)+{\alpha }_t{h}_t\left(x\right)$$

(9)

where $F_t\left(x\right)$ is the prediction at step $t$, $h_t\left(x\right)$ is the new decision tree, and ${\alpha }_t$ is the learning rate. Random Forest, on the other hand, constructed multiple decision trees during training and outputs the mode of classes or mean prediction of individual trees, reducing variance and preventing overfitting that was crucial for maintaining the robustness in noisy environments typical of VANETs.

$$Y={\displaystyle \frac{1}{N}}{\sum }_{i=1}^N{h}_i\left(x\right)$$

(10)

In Equation (10), $Y$ represents the output prediction, $N$ is the number of trees, and $h_{i\left(x\right)}$ is the prediction of the $i$-th tree. Whereas integrating these (i.e., XGBoost and Random Forest) with a Kalman Filter, which continuously estimated the state of a dynamic system amidst random noise, significantly improved the detection sensitivity by refining the predictions based on real-time updates. This setup adeptly addressed incomplete data challenges and aided in distinguishing between spoofing and non-spoofing anomalies by continuously adjusting to new data and recalibrating prediction models as more information became available:

$$x_k=x_{k-1}+K_k(z_k-H_k{x}_{k-1})$$

(11)

In Equation (11), $x_k$ is the estimated state, $z_k$ is the actual measurement, $K_k$ is the Kalman Gain, and $H_k$ is the measurement function. Herewith, the challenges such as distinguishing between types of anomalies were resolved by fine-tuning model parameters and enhancing the training dataset to include diverse spoofing scenarios to ensure the methodology’s robustness against various spoofing tactics (i.e., signal replay attacks, meaconing, signal simulation, time synchronization manipulation, encryption bypass techniques, carrier wave spoofing, code delay spoofing, signal level spoofing).

3.4. Incorporation of Attribute-Based Encryption for Enhanced Security

To further strengthen the framework security, we implemented and deployed Attribute-Based Encryption (ABE) which played a pivotal role in securing communication channels against spoofed data. ABE allowed for the encryption of data based on attributes (such as, but not limited to vehicle type; driver role; geographical location; network access level; time of access; device ID), ensuring that only entities possessing specific credentials can access the information. This capability was crucial in our established vehicular network where varying levels of data access needed to be maintained dynamically. Consequently, the ABE’s implementation process involved defining the attributes relevant to different network roles and encrypting the communication channels accordingly. This setup provided fine-grained access control, significantly mitigating the risk posed by spoofed signals attempting unauthorized access. To maintain optimum performance requirements, we implemented the ABE model in consideration of the following specifics:

(a)

A 256-bit key size was chosen to provide robust encryption while maintaining efficient processing capabilities.

(b)

The encryption payload was dynamically sized based on the specific attributes used, ensuring efficient use of bandwidth and storage.

(c)

The ABE algorithm was configured to operate with minimal rounds (i.e., 1 and 3 rounds) of computation to reduce cryptographic latency. One round was sufficient for scenarios requiring the fastest possible data processing with minimal security demands, such as in less dense traffic or controlled environments. Conversely, up to three rounds were applied in more complex or higher-risk scenarios where additional security verification was critical, such as in dense urban settings or for commercial/emergency response vehicles.

(d)

Cryptographic latency was kept minimal to ensure real-time communication was feasible, which is a crucial necessity for operational safety in dynamic vehicular environments. We used high-performance hardware (see Table 4) to make sure our ABE security system could encrypt and decrypt data quickly. This way, important navigation and operational information was sent over the network in real time without delays. We envisioned and effectively assessed that this strategic combination of optimized algorithm and hardware adaptation was crucial for maintaining real-time communication, essential for operational safety.

(e)

High availability and fault tolerance were prioritized to cope with the high mobility and variable network conditions.

Table 4. Hardware specifications for ABE’s cryptographic operations.

Component	Specification
Processors	Octa-core, >2 GHz
Memory	16GB DDR4 RAM
Storage	NVMe SSDs, 3500 MB/s throughput
Networking Components	Gigabit Ethernet, Wi-Fi 6
Cryptographic Accelerators	PCIe-based hardware security module, ABE-256 encryption at gigabit rates

Table 4. Hardware specifications for ABE’s cryptographic operations.

Component	Specification
Processors	Octa-core, >2 GHz
Memory	16GB DDR4 RAM
Storage	NVMe SSDs, 3500 MB/s throughput
Networking Components	Gigabit Ethernet, Wi-Fi 6
Cryptographic Accelerators	PCIe-based hardware security module, ABE-256 encryption at gigabit rates

3.5. Dynamic Integration of Machine Learning and Kalman Filtering for Spoofing Detection

$$Y=w_0{\sum }_{i=1}^n{w}_i{\varnothing }_i\left(x_i\right)+{\sum }_{i=1}^n{\sum }_{j=i+1}^n{w}_{ij}{\varnothing }_{ij}\left(x_i,x_j\right)+KF(z)$$

(12)

As described in Equation (12), $Y$ is the predicted outcome of whether a spoofing attack is happening, $w_0$, $w_i$, $w_{\left\{ij\right\}}$ are the weights assigned to individual and interaction terms, ${\varnothing }_i\left(x_i\right)$ and ${\varnothing }_{ij}\left(x_i,x_j\right)$ represent the features derived from data processed by XGBoost and Random Forest, and $KF\left(z\right)$ is the Kalman Filter’s output adjusting the prediction based on observed discrepancies. Our multi-layered ML approach functioned as a dynamic shield, capable of adapting to new threats in real time. Ultimately, by harnessing the strengths of XGBoost and Random Forest, we ensured high accuracy in detecting anomalies and a swift response to potential security breaches. This sophisticated system, working in tandem with the robust ABE security measures, created a formidable defense against evolving spoofing attacks. This comprehensive strategy significantly bolstered the integrity and trustworthiness of navigation and communication within implemented vehicular networks.

4. Emulation Setup and Assessment Outcome

The emulation setup was precisely designed to experiment real-world scenarios and assess the efficacy of the proposed spoofing mitigation technique (i.e., as exhibited in Figure 3). A controlled environment was established, incorporating a range of BeiDou receivers with varying levels of vulnerability to spoofing attacks. To simulate diverse attack vectors, a BeiDou signal simulator was employed, capable of generating spoofed signals with adjustable parameters such as signal strength, timing offsets, and ephemeris data manipulation. The simulator was configured to mimic both sophisticated and simplistic spoofing techniques, allowing for a comprehensive evaluation of the proposed defense mechanism. A diverse array of data types was collected during the emulations that included raw BeiDou signals, receiver measurements (pseudorange (herewith, the Pseudorange’ is the estimated measurement of the distance between a satellite and a receiver, derived from the time a signal takes to travel from the satellite to the receiver. This measurement is termed ‘pseudorange’ because it incorporates various inaccuracies, including clock discrepancies in the satellite and receiver, atmospheric interferences, and other distortive elements. Addressing these inaccuracies is essential for precise positioning and navigation. This accuracy is especially critical in researching spoofing attacks within BeiDou-enabled vehicular networks, as the reliability of pseudorange data significantly influences the success of detecting and countering spoofing activities), carrier phase (thus, the Carrier phase’ refers to the phase angle of the carrier wave used in the transmission of satellite signals. It represents the fractional part of the signal’s complete wavelength cycle at the moment of reception relative to a reference point. This phase information is crucial for enhancing the precision of GNSS measurements, enabling highly accurate distance estimations between the satellite and the receiver), signal-to-noise ratio), and navigation solutions. High-precision data logging tools were utilized to capture the subtle nuances of signal characteristics and anomalies induced by spoofing.

It is evident from Figure 3 that our emulation setup incorporated a diverse set of criteria such as link establishment requests, network connectivity checks, session logs, ad hoc terminations, critical alerts, spoof signal generation, vehicle disconnections, and isolation of malicious nodes. These components were crucial in providing a thorough examination of the projected system’s performance under conditions that mimic real-world scenarios, offering deep insights into its effectiveness across various spoofing situations. Rigorous testing of the system’s capability to initiate and maintain secure connections and quickly respond to disruptions & spoofing vulnerabilities affirmatively validated the resilience of the proposed BeiDou spoofing detection framework. During investigation, we observed that the system’s precise session tracking and rapid isolation of programmed corrupt nodes underscored its ability to protect essential communications within VANETs from complex spoofing attacks. To encounter this issue, we automated response mechanisms that were triggered upon the detection of suspicious activities to swiftly isolate the affected nodes based on predefined security rules (such as, but not limited to, block traffic from nodes exhibiting irregular signal strengths or timing discrepancies; restrict communication for nodes generating high volumes of traffic in unusually short timespans; disable connections from nodes with abrupt changes in geographical location data; automatically quarantine nodes with repeated failed authentication attempts, etc.) and algorithms, thereby minimizing potential impacts.

Herewith, the spectrum analyzer settings were optimized to ensure accurate and detailed signal analysis. The resolution bandwidth was set to a narrow value to resolve closely spaced signal components, while the internal preamplifier and attenuation were adjusted to maintain optimal signal levels for analysis. The marker bandwidth was configured to precisely measure the power and frequency of specific signal components of interest. A high-quality temperature-compensated crystal oscillator (TCXO (i.e., Trimble BD982 GNSS Receiver)) was selected for the BeiDou signal simulator to ensure the generated spoofed signals exhibited minimal frequency drift and phase noise. This was crucial for maintaining the realism of the simulated attacks and ensuring the robustness of the evaluation.

The hardware and software test setup (described in

Table 5. Hardware and software specifications emulation setup for BDS spoofing detection.

Component	Specification	Function in BDS Spoofing Detection	Justification	Settings
GNSS Receivers	Multi-frequency capability, high sensitivity	Captured BDS signals for analysis	Ensured accurate detection of authentic and spoofed signals	Frequency range: 1.2 GHz–1.6 GHz
Computing Units	High-speed processors, octa-core CPU, 16GB RAM	Processed data and ran detection algorithms	Facilitated real-time processing and analysis	Clock speed: 3.2 GHz
Software	Python 3, MATLAB, custom spoofing detection algorithms	Analyzed signal integrity and identified spoofing patterns	Allowed for flexible and powerful analytical capabilities	Version control and continuous integration enabled
Network Interfaces	Gigabit Ethernet, Wi-Fi 6	Ensured rapid data transmission within the network	Supported high data throughput and low latency	Network speed: 01 Gbps, Dual-band
Cryptographic Accelerators	ABE-256 encryption	Secured data transmission and storage	Provided strong encryption to protect sensitive data	Encryption rate: 01 Gbps
Data Storage	NVMe SSDs, 3500 MB/s throughput	Stored data securely and allowed fast access	Enhanced the overall speed and efficiency of the system	Capacity: 2 TB, Read/write speed: 3500 MB/s
Power Supply	Uninterruptible Power Supplies (UPS)	Maintained system operation during power disruptions	Ensured reliability and continuity of operations	Power output: 500 W, Efficiency: 90%
Signal Analyzers	High-resolution spectrum analyzers	Monitored and evaluated signal quality and integrity	Critical for detecting and analyzing spoofed signals	Resolution bandwidth: 1 Hz to 10 MHz
Antennas	High-gain, multi-band antennas	Received a wide range of frequencies from BeiDou satellites	Maximized the reception quality and minimized loss	Gain: 12–15 dB, Omni-directional
Cooling System		Maintained optimal temperature for system stability	Prevented thermal throttling and enhanced performance	Temperature control: Up to 45 °C

) comprised a combination of commercial COTS BeiDou receivers, SDRs, and custom-developed signal processing algorithms. The COTS receivers served as the targets of the spoofing attacks, while the SDRs were used to capture and analyze both authentic and spoofed BeiDou signals. The signal processing algorithms were implemented on a high-performance computing platform to enable real-time analysis and detection of spoofing anomalies. An overpowered attack scenario was simulated by transmitting a spoofed BeiDou signal with significantly higher power than the authentic signal. This aimed to assess the resilience of the receivers against overwhelming interference. The detection of this attack was facilitated by monitoring the C/No of the received signals, as the spoofed signal’s higher power would lead to an abnormal increase in the C/No.

To evaluate the effectiveness of the proposed mitigation technique in different operational modes, spoofing attacks were conducted on BeiDou receivers in both cold-start and tracking modes. To investigate the impact of signal distortion on spoofing detection, an overpowered attack with noise padding was simulated. The spoofed signal was intentionally corrupted with noise to obscure its characteristics and evade detection. The effectiveness of C/No monitoring in detecting this type of attack was assessed. Correlated measurements were employed to enhance the detection of spoofing attacks. By analyzing the correlation between multiple receiver measurements, such as pseudorange and carrier phase, subtle inconsistencies introduced by spoofing could be identified. This approach leveraged the redundancy in the measurements to improve the reliability of spoofing detection.

To validate the practical applicability, the proposed technique was applied to live spoofing data collected from real-world scenarios. This involved analyzing BeiDou signals captured during actual spoofing attacks to assess the effectiveness of the detection algorithm in real-time and under varying environmental conditions. Statistical processing (i.e., continuity fault tree) of multiple metrics was performed to enhance the robustness of spoofing detection. The continuity fault tree was used to model the probability of spoofing based on the occurrence of multiple anomalies in the received signals. Pseudorange residuals, which represent the difference between the measured and expected pseudoranges, were also analyzed statistically (i.e., by applying outlier detection) to identify deviations caused by spoofing. Likewise, the signal quality monitoring (SQM) metrics, such as C/No, pseudorange, and carrier phase, were utilized to assess the integrity of the received BeiDou signals. The impact of these metrics on spoofing detection was investigated, and thresholds (as described in

Table 6. Optimum SQM Thresholds.

SQM Metric	Threshold Type	Observed Optimal Threshold	Rationale
C/No	Lower Limit	45 dB-Hz	Below this, the signal is too weak for reliable tracking and was susceptible to spoofing.
Pseudorange	Deviation from Expected	10 m	Larger deviations indicated spoofing-induced errors in distance measurement.
Carrier Phase	Cycle Slip Detection	1 cycle	Cycle slips disrupted continuous phase tracking and were a sign of spoofing.

) were established to differentiate between authentic and spoofed signals based on their SQM characteristics.

Consequently, a symmetric difference metric was derived to quantify the discrepancy between the authentic and spoofed BeiDou signals. This metric was based on the comparison of multiple signal parameters, including pseudorange, carrier phase, and Doppler frequency. The symmetric difference metric provided a comprehensive measure of the signal distortion caused by spoofing. The relationship between signal power and distortion was investigated to understand how the power of the spoofed signal affects its ability to mimic the authentic signal (as represented in

Table 7. Relationship between signal power and distortion.

Spoofed Signal Power (Relative to Authentic Signal)	Distortion Level	Impact on Spoofing Effectiveness	Detection Difficulty
Low (e.g., −10 dB)	High	Less effective, easier to detect	Easier
Medium (e.g., 0 dB)	Moderate	Balance between effectiveness and detectability	Moderate
High (e.g., +10 dB)	Low	More effective, harder to detect	Harder

). This analysis provided insights into the optimal power levels for spoofing attacks and the corresponding challenges for the detection algorithm. Likewise, we also evaluated the impact of spoofing attacks on the positioning accuracy under different scenarios. This involved analyzing the position errors induced by spoofing in various vehicular environments, such as urban canyons, open highways, and underpasses. The results provided us with insights into the potential consequences of spoofing on the safety and reliability of VANETs.

Our simulation efforts were intensified by creating a navigation scenario to analyze the impact of spoofing attacks on vehicle guidance and control systems. We examined how vehicles react to altered navigation instructions to pinpoint potential risks and weaknesses. Furthermore, we investigated the simultaneous processing of spoofed and genuine BeiDou signals to develop a technique that can differentiate between them. This involved analyzing the multiple correlation peaks that can occur when both signals are present to identify and isolate the authentic signal. Practical considerations, such as distorted peaks and multipath interference, were considered during the emulation setup. To further strengthen the projected model, a localization scenario with unknown correspondence was investigated under challenging conditions. In this scenario, the receiver did not have prior knowledge of which signals were authentic and which were spoofed. The ability of the technique to correctly identify and utilize the authentic signals for localization was evaluated. The proposed spoofing mitigation technique was once again applied to various spoofing scenarios to assess their effectiveness against different attack vectors. This involved simulating a range of spoofing techniques, such as meaconing, signal replay, & data manipulation, and evaluating the ability of the techniques to detect and mitigate these attacks. Ultimately, for comparative purposes, the integrity of the navigation solution provided by the preconfigured BeiDou receivers was assessed in the presence of spoofing attacks. This involved analyzing the consistency and reliability of the position, velocity, and time information provided by the receivers under spoofed conditions.

We critically assessed emulation performance metrics such as accuracy, precision, recall, and F1-score to measure the efficacy of the implemented model. Accuracy quantified the overall effectiveness of the model in correctly identifying both spoofed and legitimate signals. It is illustrated in Equation (13):

$$Accuracy={\displaystyle \frac{TP+TN}{TP+TN+FP+FN}}$$

(13)

where $TP$ were true positives, $TN$ were true negatives, $FP$ were false positives, and $FN$ were false negatives. This metric was pivotal in assessing the general reliability of the detection system across varied test scenarios. Precision measured the model’s effectiveness in labeling an instance as positive that was truly positive and was particularly important in contexts where the cost of a false positive was high. Precision was examined as shown in Equation (14):

$$Precision={\displaystyle \frac{TP}{TP+FP}}$$

(14)

High precision in this context ensured that the system minimized interference with normal operations by not misclassifying legitimate signals as spoofing. Accordingly, Recall indicated the ability to find all actual positives, as expressed in Equation (15):

$$Recall={\displaystyle \frac{TP}{TP+FN}}$$

(15)

For the proposed BDS spoofing detection, high recall was crucial for ensuring no spoofing attacks went undetected, directly contributing to the security and safety of the navigation system. Finally, the F1-score combined precision and recall into a single metric by taking their harmonic mean, useful for comparing two classifiers, Precision and Recall, as conveyed in Equation (16):

$$F1\text{-}Score={\displaystyle \frac{Precision\times Recall}{Precision+Recall}}$$

(16)

The F1-score was vital for evaluating the balance between precision and recall, ensuring the model was not biased toward one at the expense of the other.

Table 8. Ten-fold emulation outcome of proposed methodological framework.

Emulation Iteration	Accuracy (%)	Precision (%)	Recall (%)	F1-Score (%)
1	93.2	92.4	90.3	92.5
2	94.4	91.7	92.3	93.4
3	95.3	94.2	93.2	94.2
4	92.6	92.4	90.8	91.3
5	95.1	92.6	91.3	92.4
6	94.2	93.1	92.4	92.5
7	94.6	94.6	93.3	93.5
8	94.6	93.4	91.0	92.6
9	92.5	92.8	91.3	91.9
10	93.1	94.4	93.4	94.7

validates the resilience of the detection algorithm, highlighting its consistent accuracy in identifying and mitigating spoofing attacks across diverse test runs and scenarios. These metrics offered a holistic evaluation of the model’s performance, enabling a comprehensive understanding of its strengths and weaknesses in detecting and mitigating spoofing attacks.

In contrast to traditional methods that primarily focus on GPS, as investigated by Murad et al. [8], Bethi et al. [9], Chen et al. [10], David et al. [11], Oligeri et al. [12], and Vitale et al. [13], our methodology encompasses a wider range of satellite navigation platforms, including BeiDou. This broader scope enhances the relevance and effectiveness of the proposed solution in diverse operational scenarios. The incorporation of BeiDou into the emulation assessment introduced a new layer of complexity and provided a unique perspective on spoofing defense strategies. Rigorous testing of navigation message design, cipher key updating protocols, and signature information protection mechanisms were conducted to enhance the security framework. These components were carefully evaluated through an array of performance metrics, as indicated in Figure 4. The results from this analytical approach demonstrated a marked improvement in detection capabilities, showcasing the advanced engineering and adaptation of a hybrid security solution that combined traditional cryptographic defense with cutting-edge machine learning algorithms.

Our comparative emulation setup subjected each system to a variety of attack simulations to assess their resilience and adaptability in real-time threat scenarios. The evaluation matrix provided a detailed assessment of how each methodology performed under stringent testing environments, emphasizing the superior performance of our proposed system. The proposed methodology not only exhibited higher efficacy in all key metrics but also set novel norms in the application of integrated technological solutions for satellite navigation security. Figure 5 illustrates that the proposed method stands out in detecting spoofing attacks, especially when we look at how it handles ‘Message Reception Probability’ over varying distances between the sender and the receiver. It consistently shows higher reception probabilities at increased distances, proving its effectiveness and robustness in the kind of unpredictable, real-world scenarios we often face. This is crucial because a high message reception probability, regardless of distance, means the system is less likely to fall prey to the signal degradation or disruptions that spoofing attacks can cause. The resilience of the PSAU-Defender in maintaining communication integrity, even as distances fluctuate, makes it particularly well-suited for vehicular networks, where distances between senders and receivers can change quickly and unpredictably. This ability to keep connection probabilities high in dynamic environments directly enhanced the reliability and security of emulated VANETs.

4.1. Limitations of PSAU-Defender

It is evident that like any state-of-art system, the projected methodology also inherently has limitations because it is designed with specific objectives, constraints, and resources in mind, which cannot cover all possible scenarios or conditions. Thus, during emulation we anticipated and observed the following limitations:

(a)

We acknowledge that our investigation relied heavily on a COTS BeiDou receiver that did not fully represent the diversity of hardware used in real-world scenarios. This limitation could affect the generalizability of the findings, as performance nuances specific to other hardware types are not explored.

(b)

While the methodology incorporates advanced cryptographic and machine learning techniques, the computational demand and hardware requirements may not scale efficiently in larger or more complex VANET environments. This could impact the practical deployment of the proposed solutions in diverse operational settings.

(c)

In some iterations, the emulation setup did not adequately account for various environmental and physical interferences that can affect signal transmission, such as urban canyon effects, multipath reflections, and atmospheric conditions. These factors may partially influence the performance of spoofing detection systems.

(d)

We acknowledge the possibility that in metropolitan VANETs the use of complex XGBoost and Random Forest with Kalman Filtering may pose a risk of overfitting that can lead to higher error rates in unfamiliar or evolving attack scenarios.

(e)

Although the implementation of ABE has improved data security, the management of encryption and decryption processes within varied and distributed VANET environments (i.e., large scale metropolitan networks) may lead to overheads and raise concerns about data privacy, particularly in cases of node compromise or insider threats. Addressing this limitation will be a focus of our future research efforts.

4.2. Electric Vehicles Security Enhancement

It is clear from the methodology and results we have presented that our approach to detecting BeiDou spoofing attacks in VANETs is especially relevant for electric vehicles (EVs). This aligns well with the broader goal of boosting security in advanced transportation systems. Electric vehicles are integral to modern smart transportation networks which benefit significantly from secure and reliable navigational data to optimize route planning, energy management, and automated driving functions. By fortifying the integrity of satellite navigation signals against spoofing attacks, our framework directly supports the safety and efficiency of electric autonomous vehicles, contributing to the stability of vehicle-to-vehicle and vehicle-to-infrastructure communications. This is essential not only for the individual vehicle’s operational reliability but also enhances the overarching smart grid and charging infrastructure by ensuring precise and trustworthy location data. Thus, we are confident in claiming that our research underpins critical facets of electric vehicle management, including energy and thermal management systems that rely on accurate, spoof-free navigational inputs to function optimally within the connected ecosystems of modern electric and hybrid vehicles.

5. Conclusions

This study presented a robust methodology for detecting and managing BeiDou BDS spoofing in VANETs and evaluated its efficacy through comprehensive emulation setups. The research demonstrated significant advancements in spoofing detection technologies by utilizing hybrid machine learning technique (i.e., XGBoost and Random Forest with Kalman Filter) and advanced cryptographic methods (i.e., Attribute-Based Encryption) to enhance the security and reliability of navigation systems in dynamic vehicular environments. Despite achieving promising results, the study acknowledges limitations such as potential scalability issues and the need for higher processing capabilities in the emulated application. Our emulation-driven findings are crucial for the ongoing development of secure vehicular communications and provide a sturdy foundation for further research. Future studies could focus on optimizing the computational efficiency of the detection systems and expanding the robustness of the security measures to cover newer spoofing techniques, potentially integrating broader machine learning models and more sophisticated cryptographic solutions to better mimic real-world conditions and further improve system resilience.