Ai-driven Post-Quantum Cryptography Enhances Cyber-Resilient V2X Communication in Transportation Cyber-Physical Systems

Transportation Cyber-Physical Systems, which seamlessly integrate vehicles and infrastructure, increasingly rely on secure communication to ensure safety and efficiency, but face growing threats from quantum computing. Akid Abrar, Sagar Dasgupta, and Mizanur Rahman, all from the University of Alabama, alongside Ahmad Alsharif, investigate how artificial intelligence can bolster Post-Quantum Cryptography to protect these critical systems. Their work addresses the vulnerability of current cryptographic methods to future quantum attacks, and demonstrates how AI can optimise algorithm selection and resource allocation for enhanced security. By combining the resilience of Post-Quantum Cryptography with the adaptive capabilities of artificial intelligence, this research establishes a pathway towards cyber-resilient communication within the evolving landscape of transportation networks.

The work focuses on transitioning to PQC in the face of emerging quantum computing threats, specifically within connected and autonomous vehicles. A significant emphasis is placed on practical implementation challenges and the need for hybrid approaches that combine existing and future cryptographic methods. The research covers several key areas, including the implementation of PQC algorithms, such as Kyber, and the development of hardware solutions for resource-constrained devices found in vehicles.

Scientists are investigating how to combine classical cryptography with PQC to ensure a smooth transition and maintain compatibility with existing systems. A central focus is securing Vehicle-to-Everything (V2X) communication channels against evolving cybersecurity threats, utilizing intrusion detection systems powered by machine learning to identify malicious activity. Artificial intelligence, particularly machine learning and reinforcement learning, is proposed as a critical enabler for PQC. Researchers are exploring how reinforcement learning can optimize PQC algorithm implementations, adapt security measures dynamically, and efficiently allocate resources in cloud and edge computing environments supporting V2X communication.

AI is also being investigated for mitigating side-channel attacks, which exploit physical characteristics of cryptographic devices. Practical implementation challenges, such as limited processing power, memory, and energy in embedded systems, are driving innovation in efficient and optimized PQC algorithms and security protocols. Lattice-based cryptography is a prominent family of PQC algorithms being investigated, and organizations like NIST are driving the standardization process. The GSM Association is developing guidelines for PQC in telecom applications, and collaborations between companies like SoftBank and SandboxAQ are exploring hybrid PQC solutions. The study addresses escalating cyber threats facing TCPS, which rely on Vehicle-to-Everything (V2X) communication for safety and efficiency. Researchers recognized that existing cryptographic methods are vulnerable to attacks from quantum computing, necessitating a transition to PQC algorithms. The work centers on enhancing the effectiveness of PQC through AI-driven optimization. Scientists developed methods to intelligently select and allocate PQC algorithms, adapting to evolving threats in real-time.

This approach moves beyond static security measures, enabling dynamic adjustments to cryptographic protocols based on observed vulnerabilities and system performance. The team investigated vulnerabilities in V2X communications, identifying critical data points, including geolocation, trajectory data, authentication credentials, and firmware updates, that require robust protection. The study highlights the significant safety benefits of V2V and V2I communication systems, estimating they could prevent or mitigate up to 80% of crashes not involving impaired drivers. Independent modeling suggests injury crashes could fall by 43 to 55%, and fatal crashes by 31 to 37%, with full fleet penetration. Scientists have shown that while PQC algorithms like Kyber KEM and Dilithium signatures are viable for securing vehicle-to-everything (V2X) communications, simply replacing existing cryptographic methods presents challenges in meeting the strict real-time constraints of TCPS. Experiments revealed that incorporating Kyber into a TCPS setting performed well on high-bandwidth wired links, but on latency-sensitive wireless links, the increased key size and computational demands introduced difficulties in consistently meeting the 100ms verification time required for safety applications. To address these limitations, researchers developed AI-driven frameworks capable of dynamically selecting the most appropriate cryptographic algorithm based on context.

A key achievement is the demonstration of a KNN-based controller that can switch between AES-only encryption and a hybrid ECC-AES encryption based on data sensitivity, reducing overhead while preserving confidentiality. Furthermore, scientists showcased reinforcement Q-learning to scale encryption levels in wireless-sensor networks, achieving a 30% reduction in energy consumption without sacrificing packet-delivery ratio. A machine learning model was trained on parameters like message size, required latency, node CPU load, and channel conditions to determine whether to use Dilithium or Falcon for signatures, or a hybrid RSA+Kyber handshake versus a pure Kyber handshake for key exchange. This AI-driven selection enables true cryptographic agility, allowing the TCPS to continuously adapt to emerging threats and maintain optimal performance across heterogeneous devices and network conditions. The study confirms that safety-critical communications within TCPS, including vehicle-to-vehicle and vehicle-to-infrastructure exchanges, face threats from eavesdropping, spoofing, and data manipulation due to weaknesses in currently used cryptographic methods. These existing schemes, such as RSA and ECDSA, are increasingly at risk from algorithms like Shor’s and Grover’s, which could compromise the security foundations of connected transportation. Trials on traffic controllers and onboard units indicate that, with careful parameter tuning, these schemes can meet the strict latency requirements for safety-critical applications. However, the study acknowledges that increased computational load and packet volume may pose challenges in high-density traffic scenarios, necessitating a staged implementation approach. Furthermore, the research suggests that artificial intelligence (AI) can play a crucial role in optimizing PQC adoption, enabling systems to dynamically adjust cryptographic parameters and offload computations to enhance performance and resilience.