Research Projects

Lead Principal Investigator(s): Satish Ukkusuri (Purdue University)
Co-Principal Investigator(s): Alvaro Cardenas (The University of California, Santa Cruz), Daniel Fremont (The University of California, Santa Cruz), Leilani Gilpin (The University of California, Santa Cruz), Gurcan Comert (Benedict College), Mansoureh Jeihani (Morgan State University), Mashrur Ronnie Chowdhury (Clemson University), M Sabbir Salek (Clemson University)
Research Project Funding: Federal $266,304.78; Cost-share $266,838.78
Project Start and End Date: January 1, 2024, to December 31, 2024
Project Status: Report in Progress

Project Description: Many states and local administrators have vowed to advance advanced transportation systems by enhancing autonomy and connectivity. While integrating new technologies and algorithms holds promise in promoting efficiency and safety, it also introduces vulnerabilities. Previous research has demonstrated viable attacks on connected and autonomous vehicles (CAVs), such as GPS spoofing and tactics involving the manipulation of traffic signals. However, most studies are based on small-scale scenarios (e.g., one vehicle, one intersection, or one link), which can only reflect the local and limited impact of the attacks. To comprehensively evaluate the threats associated with cyberattacks against CAVs, and to see whether specific defense mechanisms effectively address a threat, a faithful testbed capable of handling multi-scale system dynamics is needed.

The proposed project aims to develop a sophisticated simulation testbed capable of assessing the multi-scale impact of cyber-attacks against CAV fleets. Unlike existing testbeds, our project will adopt a co-simulation framework to model multi-scale system dynamics from V2X communication, vehicle maneuvering, and car-following to vehicle scheduling, routing, and network-level cascading congestion effects. Ultimately, this project aims to construct a reliable environment that can serve as a foundational platform for future cybersecurity studies.

USDOT Priorities: The project supports USDOT priorities and the RD&T strategic goals by:

• Safety: The testbed enables monitoring of existing vulnerabilities, assessing their risks, and testing diverse defense algorithms. This contributes to building a safer transportation system for all people.
• Economic Strength and Global Competitiveness: The testbed plays an essential role in developing more secure and reliable connected and autonomous vehicle applications, strengthening their global competitiveness. This contributes to the fostering of an inclusive and sustainable economy.
• Transformation: The testbed provides tools for addressing cybersecurity challenges for future connected/autonomous vehicle applications. This contributes to the deployment of new transportation applications, driving transformative advancements in the practical field.

The project engages in breakthrough, advanced, or transformative research by:

• Holistic Vulnerability Assessment: Our project will simulate the large-scale impact of cyberattacks in connected and autonomous vehicle (CAV) systems, providing a comprehensive view of potential weak points and risks across the network.
• Practical Attack Validation: By verifying existing attacks on CAV fleets, we bridge the gap between theory and practice, offering tangible insights into the real-world impact of these threats.
• Unveiling Cascading Attack Effects: The testbed will capture and analyze the cascading effects of cyber-attacks, shedding light on how disruptions can propagate within road networks, shedding light on a more resilient and secure transportation infrastructure design.
• Enhancing Traffic Flow Understanding: Our project will model compromised traffic flow under cyber-attacks, allowing us to accurately predict and mitigate operational disruptions, ultimately contributing to safer and more efficient traffic management.

Outputs:

• A High-fidelity Testbed for Cyberattacks: This project will develop a co-simulation testbed with three components (cloud controller, scenario generator, and a high-fidelity traffic simulator) based on state-of-the-art technologies. Our code and data will be open-source and documented to make them available for examination and future studies.
• Simulation Language for Cyberattacks: This project will create a solid mechanism that allows us to define diverse cyberattack scenarios targeting CAVs formally. Through this mechanism, we can specify various cyber threats, ranging from GPS spoofing to malware intrusions, to assess their potential impact on CAV fleets reliably.
• Testing of Typical Cyberattacks: This project will test typical attacks on CAV fleets that are reported in the literature and provide crucial insights into their potential cascading impacts on road networks, helping to forge a better understanding of the potential risks of CAVs among policymakers and the public.

Outcomes/Impacts:

• This project will lay the foundation for expansive future endeavors in transportation cybersecurity research. Our vision encompasses the initial testbed's efficacy and its potential for growth and adaptation. To this end, we envision collaborative projects with research partners that leverage our testbed's capabilities to explore new dimensions of cybersecurity in intelligent transportation systems (ITS).
• This project will provide information about the large-scale impact of cyberattacks on road networks. The findings can be integrated into practical applications such as data collection and communication standards and regulations.
• A crucial aspect of this project involves a commitment to regular updates and enhancements across all testbed components. This project aims to foster a community to continuously refine and elevate the testbed's capabilities to address evolving research imperatives and the ever-changing landscape of cybersecurity threats in CAV applications.

Principal Investigator(s): Mohammadhadi Amini (Florida International University)
Co-Principal Investigator(s): Farhad Shirani Chaharsooghi (Florida International University), Kemal Akkaya (Florida International University), Selcuk Uluagac (Florida International University), Mansoureh Jeihani (Morgan State University)
Research Project Funding: Federal $150,000; Cost-share $150,121
Project Start and End Date: January 1, 2024, to December 31, 2024
Project Status: Report in Progress

Project Description: In this project, we aim to deploy, integrate, and validate privacy-preserving and secure learning solutions for CAVs. Our proposed solution includes four major goals:

1. An integrated anomaly detection technique to detect and isolate backdoor attacks in FL settings for connected and autonomous vehicles (CAVs);
2. a hybrid approach that maintains concrete security against backdoor attacks in CAV applications;
3. a privacy preservation mechanism to ensure CAV data is protected against data leakage,
4. training proposed learning models using real-world and synthetic CAV data for assessment and validation purposes.

These four goals will mainly contribute to the “Security and Resiliency” of intelligent transportation systems while ensuring “Data Privacy,” which is aligned with USDOT's goals to secure transportation systems and TraCR’s vision. This project develops a distributed learning architecture to serve as a platform for future projects. For example, it can be used to develop and evaluate other privacy-preserving techniques for intelligent transportation systems use cases.

USDOT Priorities: This project will facilitate the integration of autonomous transportation technologies by enabling more secure and privacy-preserving solutions. In summary, it will achieve these goals by:

• Ensuring cybersecurity of autonomous transportation systems while deploying efficient and robust federated learning algorithms for image recognition
• Integrating privacy-preserving learning algorithms that are robust against privacy leakages to protect the identification of autonomous vehicle drivers and passengers in CAV applications
• Detecting and isolating backdoor cyber-attacks for CAV applications
• Advancing the security and privacy of autonomous transportation systems, which can facilitate widespread integration of CAVs
• Consensus-based federated learning algorithms for adversarial robustness in the presence of targeted backdoor attacks in CAV applications

Further, it contributes to the body of knowledge in integrating the security and privacy protection mechanisms and tailoring them towards connected autonomous vehicle applications. One of the promising solutions for efficient decision-making in CAVs is federated learning, a distributed machine learning solution that enables local model training and eliminates the need for sharing each agent’s (vehicle’s) data with a central server. This project aims to design and integrate federated learning (FL) models and algorithms for privacy-preserving image recognition for detecting and isolating backdoor attacks in autonomous transportation systems. Cyber-attacks can seriously threaten the privacy and security of FL systems. Based on the attacker's goal, cyber-attacks in FL are divided into targeted and untargeted attacks. The targeted attacker's goal is to manipulate the specific subtask of the learning process, making it difficult to identify the attack. A backdoor attack is a category of targeted attack that occurs when an attacker manipulates specific input data of an edge device. More precisely, the attacker can manipulate the sensor measurements at the edge device, such as the images captured by a camera. This project specifically focuses on detecting backdoor attacks on images in CAVs. If this attack is not detected and isolated from the systems, it leads to an incorrect understanding of the road environment. It can cause detrimental impacts by leading to accidents or misunderstanding the traffic lights by individual vehicles. In addition to adversely affecting the individual cars' accuracy of decision-making processes at each of the autonomous vehicles, such attacks may also negatively impact the global consensus of CAVs, with potentially catastrophic consequences.

Outputs: The project outcomes will be delivered every quarter aligned with the four specific research and development tasks elaborated below:

Task 1: Problem Formulation and Developing Architecture of the Hybrid Algorithm for CAVs
Deliverables: Progress report; pseudocode and formulation of problem; presentation/slides
Task 2: Preaggregation Similarity Measurement for Attack Isolation in CAVs
Deliverables: Progress report; source code of implementation and validation; presentation/slides
Task 3: Privacy-aware Operations of CAVs
Deliverables: Progress report; source code of implementation and validation; presentation/slides
Task 4: Validation of Proposed Hybrid Model Using Benchmark and Real-world CAV Dataset
Deliverables: Final Technical Report; source code of implementation and validation using multiple datasets; presentation/slides; Details of Transition to Practice Activities

Outcomes/Impacts: The impact of this project is on securing federated machine learning algorithms with an emphasis on CAV applications that will enable more secure and privacy-preserving integration of autonomous vehicles. It will integrate and tailor secure and privacy-preserving mechanisms specifically for CAV-related applications.

The project team has solid experience in commercialization and transition to practice activities. We will also benefit from StartUP FIU as a university-wide innovation hub that fosters and develops entrepreneurship and innovation to help our students, researchers, and community connect, contribute to, and thrive in today’s fast-changing world. Further, the MSU team has multiple patents related to autonomous transportation systems, notably a testbed for CAVs. Team members from both institutions will start developing a transition plan starting the third quarter of the project. The goal of this plan is to outline the next steps, future works, remaining gaps that could lead to future research projects at TraCR, and possible commercialization plans (including identifying potential industry partners).

Lead Principal Investigator(s): Yiheng Feng (Purdue University)
Co-Principal Investigator(s): Satish Ukkusuri (Purdue University), M. Hadi Amini (Florida International University), Kemal Akkaya (Florida International University)
Research Project Funding: Federal $125,000; Cost-share $125,500
Project Start and End Date: January 1, 2024, to December 31, 2024
Project Status: Report in Progress

Project Description: In this project, we will build a multi-resolution simulation platform to test and evaluate transportation system cybersecurity. The simulation environment will build on an existing open-source co-simulation environment for cooperative driving automation (CDA) developed by the FHWA. Based on the co-simulation environment, we will develop APIs to support various attack scenarios, including sensor attacks, data spoofing attacks, infrastructure attacks, vehicle-level attacks, and network-level attacks. Further, we will investigate the impact of these attacks on V2X infrastructure applications, machine learning algorithms, and network routing applications.

USDOT Priorities: This project supports USDOT’s research priorities in “reducing transportation cybersecurity risks.” It also promotes mobility and safety since cybersecurity is closely coupled with them in the context of CAV ecosystems. The project will advance state-of-the-art research in security analysis and mitigation methods in the CAV ecosystem at multiple levels, from vehicle and intersection to network.

Outputs: The main output will be a baseline simulation environment that supports simulating vehicle level, intersection level, and network level transportation applications. Several cyber-attack interfaces (or APIs) will be provided to the users to integrate their attack models. Meanwhile, several example attacks and associated impacts will be generated.

Outcomes/Impacts: The expected outcomes will result in publications in top transportation and security journals and conferences, enhance the understanding of cybersecurity risks in the CAV ecosystem and provide insights for researchers to develop further mitigating strategies and agencies to make informed policy decisions.

Completed

Lead Principal Investigator(s): Z. Berkay Celik (Purdue University)
Co-Principal Investigator(s): Daniel Fremont (The University of California, Santa Cruz), Satish Ukkusuri (Purdue University), Alvaro Cardenas (The University of California, Santa Cruz)
Research Project Funding: Federal $125,231.78; Cost-share $126,619.78
Project Start and End Date: January 1, 2024, to December 31, 2024
Project Status: Report in Progress

Project Description: Autonomous Driving (AD) vehicles must interact and respond in real-time to multiple sensor signals, indicating how other autonomous robots, targets, and the environment behave near the ego vehicle. While autonomous vehicle (AV) developers tend to generate numerous test cases in simulations to detect problems, to our best knowledge, they are not testing for malicious physical interactions from attackers, such as placing emergency cones in the hood of an AV or driving maneuvers that nearby human vehicle drivers or other AV manufacturers can create. For example, a hostile driving maneuver causing the victim vehicle to crash (while the malicious driver does not crash) can be identified by malicious actors and then spread and reproduced by multiple people worldwide, causing traffic accidents on vehicles with vulnerable AD stacks.

Recently, TraCR members of UCSC and Purdue have introduced two frameworks to explore the practicability of adversarial physical conditions in real-world environments. They focused on adversarial driving maneuvers, a new class of physical attack against AD software. Here, the attacker aims to find a (plausible) trajectory near the victim's vehicle to cause it to behave unintendedly, such as crashing or driving off the road.

The frameworks proposed by UCSC and Purdue differ in their assumptions about the attacker and the target AV software components. However, both provide an overview of the challenges, a means of discovering adversarial driving maneuvers in practice, and potential solutions to defend against them. While both frameworks have been shown, to some extent, to be effective in discovering adversarial driving maneuvers against a variety of AD software, the research on adversarial driving maneuvers is still in its early stages. In this proposal, we will study the weaknesses and strengths of both frameworks. Guided by our findings, we will explore creating a unified framework leveraging the best ideas from each university and explore rigorous measures of adversarial maneuvers for building a safe and secure AD software stack.

USDOT Priorities: The project aligns with key USDOT priorities/RD&T strategic goals, including:

• Safety: The project aims to enhance the safety of autonomous vehicles by investigating and mitigating adversarial physical attacks. This directly addresses the USDOT's strategic goal of making our transportation system safer for all people.
• Economic Strength and Global Competitiveness: Developing secure and reliable autonomous vehicles will contribute to a more efficient and competitive transportation system, supporting the USDOT's strategic goal of growing an inclusive and sustainable economy.
• Equity: By ensuring the safety and security of autonomous vehicles, the project will promote equitable access to transportation, aligning with the USDOT's strategic goal of reducing inequities in transportation systems.

The project engages in breakthrough, advanced, or transformative research by:

• Exploring a new class of physical attacks against autonomous vehicles: The project focuses on adversarial driving maneuvers, a relatively new and underexplored threat to autonomous vehicles.
• Developing a unified framework for detecting adversarial driving maneuvers: The project aims to combine the strengths of existing frameworks into a single, comprehensive approach.
• Investigating rigorous measures for evaluating adversarial maneuvers: The project seeks to establish standardized methods for assessing the severity and effectiveness of adversarial attacks.

Outputs:

• Developing New Security Tools: Our research will lead to developing new security tools and frameworks that can be used to protect AVs from attacks. These tools could be commercialized and used by AV companies and automakers. Our code, data, and models will be open-source and packaged to make them useful to our team and external researchers.
• New Security Standards and Regulations: Our research could inform the development of new security standards and regulations for AVs. These standards and regulations can ensure that AVs are developed and deployed securely.
• Educating the Public: Our research could be used to inform the public about the risks and challenges of AV security, helping build public trust in AV and accelerate their adoption.

We expect our results will lead to more extensive research projects with diverse partners and proposed technology transfer plans.

• Industry: AV companies, automakers, and other technology companies are interested in improving the security of their vehicles. We will partner with these companies (e.g., Toyota Research Institute of North America (TRINA), Waymo, Ford, and GM) by leveraging TRaCR contacts to explore adversarial maneuvers and develop solutions to mitigate those vulnerabilities.
• Government Agencies: Government agencies, such as the Department of Transportation and the National Highway Traffic Safety Administration, are interested in improving the safety of AVs. We plan to partner with these agencies to research the broader implications of AV security, such as developing new regulations and standards.
• Academia: We will explore partnering with researchers at other universities and research institutions (e.g., the University of Michigan and the University of California, Irvine) to inform them of our findings, share ideas, and explore collaboration opportunities.

Outcomes/Impacts: The research output of this project could lead to several changes to the transportation system.

• Improved Safety: The project's output could help make autonomous vehicles safer for all users by reducing the risk of successful adversarial attacks.
• Increased Reliability: By making AVs more resilient to adversarial attacks, the project's output could help to improve their reliability and availability.
• Reduced Costs: By preventing damage to AVs and infrastructure caused by adversarial attacks, the project's output could help to reduce the overall cost of operating and maintaining autonomous transportation systems.

Principal Investigator(s): Murat Kantarcioglu (The University of Texas at Dallas)
Co-Principal Investigator(s): Latifur Khan (The University of Texas at Dallas), Bhavani Thuraisingham (The University of Texas at Dallas), Alvaro Cardenas (The University of California, Santa Cruz), Gurcan Comert (Benedict College)
Research Project Funding: Federal $124,124,78; Cost-share $124,264.78
Project Start and End Date: January 1, 2024, to December 31, 2024
Project Status: Report in Progress

Project Description: The large-scale user data collection has enabled various new services to improve transportation services with crowdsourced vehicle routing applications or public transit metrics. This fine-grained collection of user data benefits society but raises privacy concerns as attackers can obtain location and trajectory data from various users. On the other hand, researchers need realistic data to perform experiments that can improve the efficiency of transportation systems; however, the sensitive nature of this data, which can include personally identifiable information, often prevents it from being openly shared or utilized for broader research and public benefit. Thus, while transportation data holds significant potential for improving infrastructure and services, privacy considerations create a barrier that must be carefully navigated. Privacy-preserving synthetic data generation represents a promising avenue for addressing the challenges of sharing transportation data. Privacy-preserving synthetic data can be engineered to retain the essential characteristics and statistical properties of the original dataset while removing or altering information that could compromise individual privacy.

This approach enables researchers, policymakers, and urban planners to gain valuable insights into transportation patterns, traffic congestion, and infrastructure needs without violating individual privacy. In this way, privacy-preserving synthetic data can be a powerful tool for enhancing transportation systems while providing privacy protections. Due to the above-mentioned reasons, in this project, we will work on developing novel privacy-preserving synthetic transportation data generation.

USDOT Priorities: The DOT research strategic plan mentions “Established and routinely updated cybersecurity and privacy standards minimize cyber-risks and maintain privacy.” as one of the important and desired outcomes for the future transportation system systems. In addition, the DOT research strategic plan mentions that “The Department is committed to supporting public sector experimentation, sharing insights, and embracing open data and transparency while protecting privacy.” Along with these goals mentioned in the DOT RD&T plan, this project will provide tools and techniques to enhance data privacy while sharing transportation data. Our overall research goal is to create new privacy-preserving data generators that output synthetic data conditioned on spatiotemporal parameters, such as time and location, and to develop privacy-preserving data analytics tools for TCPSS that can effectively leverage these synthetic datasets for critical decision-making.

Outputs: We will work on four different tasks to create privacy-preserving synthetic data generation tools tailored for transportation applications. In task 1, we will analyze different data generation models’ utility concerning transportation planning and usage and develop utility measures to compare different synthetic data generation techniques. For example, we will measure the accuracy of generated data in predicting accidents in a specified location. In task 2, we will explore the impact of existing attacks in inferring sensitive information from synthetic transportation data. We will develop new attacks to understand the privacy protections provided by the existing techniques. In task 3, we will develop a novel privacy-preserving transportation data generation technique that considers the impact of important events, such as accidents, on the underlying data.

Furthermore, we will investigate tradeoffs in the performance, privacy, and scalability of the proposed data-driven approaches and tools and smart data fusion of various synthetic datasets and adaptive technologies to ensure transportation application adaptability and robustness even if synthetic data are used. In task 4, we will develop defense models against potential backdoor attacks that might cause our generated models to yield faulty outputs. This will make our model more robust and improve the quality of the generated synthetic data. In task 5, we will develop techniques to allow decision-makers to generate synthetic data sets to understand the impact of different policy decisions.

Outcomes/Impacts: One important novelty of our privacy-preserving synthetic data generation technique would be generating data for simulating different events during certain times and locations. For example, we would be able to generate data for scenarios with a 10% percent decrease in accidents in a certain neighborhood. Similarly, the synthetic data generation algorithm could simulate what happens if the number of events increases in a certain location and time. Leveraging our synthetic data generation algorithm to generate synthetic data under changed conditions, our tool will allow policymakers to use the generated data to analyze the overall impact on a given transportation system. For example, for the newly generated synthetic data, we plan to compute aggregate statistics such as average travel time from location x to location y on Monday mornings and then compute the same statistics under the assumed change. This project will develop different data analytics tools and techniques so that important statistics can be efficiently computed for different hypothetical scenarios. This will allow the decision-makers to simulate the impact of different policy decisions much more easily.

In summary, we expect our project to have the following outcomes/impacts:

• Understand privacy attacks against existing synthetic transportation data generation techniques.
• Develop a toolbox that includes major synthetic data generation techniques.
• Develop novel privacy-preserving synthetic data generation techniques for furnishing transportation decision-makers with privacy-preserving synthetic data that match real-world traffic flows for pre-specified locations, seasons, times of day, and populations.
• Provide decision-makers with new data analytics tools to analyze synthetic data to answer critical transportation planning questions.

Lead Principal Investigator(s): Cihang Xie (The University of California, Santa Cruz)
Co-Principal Investigator(s): Alvaro Cardenas (The University of California, Santa Cruz), Murat Kantarcioglu (The University of Texas at Dallas)
Research Project Funding: Federal $56,179.78; Cost-share $56,212.78
Project Start and End Date: January 1, 2024, to December 31, 2024
Project Status: Report in Progress

Project Description: The rise of autonomous vehicles (AVs) is transforming the transportation sector, potentially enhancing road safety, optimizing traffic flow, and bringing about a more sustainable future. Central to this revolution lies two interlinked technological keystones: integrating advanced sensor systems and applying cutting-edge machine-learning techniques. Specifically, the fusion of high-resolution imagery from cameras and the depth precision of Light Detection and Ranging (LiDAR) sensors equips AVs with an unparalleled perceptual prowess, allowing AVs to capture a holistic, 360-degree spatial awareness of their surroundings. Subsequently, machine learning algorithms transform the collected sensor data into actionable insights, empowering the vehicle to make accurate and informed driving decisions.

While machine learning algorithms help autonomous driving systems exhibit remarkable capabilities in recognizing patterns and making decisions, they also harbor an Achilles' heel known as adversarial vulnerability. It has been previously shown that attacks can mislead the vehicle into misrecognizing traffic signs, misjudging obstacles, or misinterpreting road conditions. Such vulnerabilities pose profound safety risks, as malicious actors could exploit them to induce unintended behaviors in AVs, potentially leading to hazardous situations on the road. As self-driving technology accelerates, understanding and mitigating these adversarial vulnerabilities becomes paramount to ensure the safety, reliability, and public trust in autonomous transportation.

This project aims to provide a multi-dimensional security analysis for advanced autonomous driving systems. Specifically, we pivot our investigation toward the Bird's Eye View (BEV) — a cutting-edge 3D perception system now gaining traction in real-world self-driving systems. The perceptual capabilities of this considered system will be further enhanced via the integration with LiDAR signal. It is noteworthy that despite its growing prevalence in modern AVs, the BEV system remains a relatively untapped area in adversarial machine learning research. Moreover, beyond merely focusing on fooling AVs’ perception system to recognize objects of interest as in existing studies wrongly, this project orients towards adversarial scenarios where attackers can induce tangible, real-world disruptions — such as instigating traffic congestions or triggering vehicular collisions — especially when interacting with other dynamic agents like vehicles or pedestrians.

USDOT Priorities: This project is aligned with the USDOT priorities defined by the National Intelligent Transportation Systems Reference Architecture; particularly, it focuses on the topic VS01: Autonomous Vehicle Safety Systems. This project will focus on developing safe and secure vehicle perception by designing attack-resilient cameras and LiDAR systems to identify safe pedestrians, vehicles, and other objects that may cause an accident. The project will solve advanced research challenges in adversarial machine learning when applied to sophisticated new sensors.

Outputs: This project raises a new research problem and is expected to provide new insights into the security of camera-LiDAR fusion systems used in AVs. Specifically, we expect to provide,

1. New adversarial techniques designed to holistically test and improve AV systems' recognition and decision-making under attack scenarios.
2. Development of co-simulation environments to assess the effectiveness of adversarial attacks in a diverse but controlled setting.

Outcomes/Impacts: This proposal is expected to be substantial in various aspects of the transportation system, particularly regarding AV:

1. Safety Improvements: The project is likely to enhance road safety significantly --- the development of robust defense mechanisms against attacks that could lead to the misrecognition of traffic signs or other vehicles will reduce the risk of accidents.
2. Industry Standards and Practices: The project's outputs may lead to the development of new industry standards for integrating cameras and LiDAR in AV systems. As such, it could change the practice of how AVs are designed, tested, and certified.

Lead Principal Investigator(s): Steven Jones (The University of Alabama Tuscaloosa)
Co-Principal Investigator(s): Mizanur Rahman (The University of Alabama Tuscaloosa), Trayce Hockstad (The University of Alabama Tuscaloosa), Latifur Khan (The University of Texas at Dallas), Mashrur Ronnie Chowdhury (Clemson University), M Sabbir Salek (Clemson University)
Research Project Funding: Federal $153,156; Cost-share $153,640
Project Start and End Date: January 1, 2024, to December 31, 2024
Project Status: Report in Progress

Project Description: Rapidly evolving advanced transportation systems rely on computing and communications technologies to integrate and optimize our systems for moving goods and people while focusing on equitably advancing society. Transformative technologies include autonomous vehicles (AVs), vehicle-to-vehicle (V2V) and vehicle-to-infrastructure (V2I) communications, as well as hardware and software that enable the collection, storage, transfer, processing, analysis, and responses based on enormous amounts of data in real-time. This is intended to reduce congestion, delays, crashes, fuel consumption, emissions, financial costs, social inequities, and more. However, the more we rely on automation and connectivity, the more we give malicious actors unprecedented opportunities to steal data, invade privacy, demand ransom, generate misinformation, and malign the systems on which our lives, prosperity, and security depend. Although the need for regulatory and enforcement measures is dire, no single federal law or federal regulatory framework governs cybersecurity or data privacy focusing on transportation in the United States.

Innovative but legally unprecedented technological advances are creating policy issues for legislative and regulatory bodies in a world of automated mobility. These include problems surrounding the amount, nature, and potential exploitation of data collected from connected transportation systems. Perhaps most concerning, current cybersecurity regulations overwhelmingly fail to require or even encourage, the use of machine learning and predictive analysis to understand privacy threats, cyberattacks, and data theft. Instead, policymakers have focused predominantly on ex-post litigation-based remedies for consumers harmed through cybersecurity breaches. These remedies, however, are often out of reach for marginalized sectors of the population.

The unregulated use of these technologies can even raise equity and discrimination issues. The project attempts to answer: (i) what federal and/or state agencies are responsible for governing cybersecurity practices in the U.S., including risk assessment, preventative measures, detection of breaches, and remedial enforcement; and (ii) how do industry experts assess the greatest risks/threats to ensuring cybersecurity in the transportation sector? The results of these two reviews will then be analyzed using natural language processing methods to identify consistencies and gaps in the nation's cybersecurity policy and what the industry indicates it should have. Finally, this analysis will be used to develop a policy guidance document to share with stakeholders who wish to develop and implement effective cybersecurity legislation and regulatory governance.

Thus, the overarching goal of this project is to perform a nationwide survey of existing federal and state cybersecurity and privacy regulatory measures and analyze that legislative landscape, considering identified risks and threats to the transportation industry. Specifically, the objectives of this project are to (i) identify and analyze the gaps that exist in the U.S. cybersecurity regulatory schematic as applied to transportation law and policy and (ii) develop a policy guidance document and/or toolkit to assist interested stakeholders in constructing and implementing effect transportation cybersecurity measures.

USDOT Priorities: This project will address several of USDOT’s strategic goals, including:

1. Serving all citizens, particularly those from underserved backgrounds and rural areas (“Equity”);
2. Improving the safety of urban, rural and underserved communities (“Safety”), and
3. Ensuring safe and secure movements of people in rural and underserved communities (“Economic Strength and Global Competitiveness,” “Organizational Excellence”).

In addition, this project will directly focus on the TraCR’s Research Thrust 3, “Society and Environment.”

Outputs:

• Develop and apply methods of large language model (LLM) based on natural language processing (NLP) overlay to analyze consistencies and gaps between existing legislation (whether state or federal) and the identified needs of the transportation industry.
• Develop a broadly applicable policy guidance document to be shared with interested transportation research and policy-making participants, which will assist in developing and implementing effective cybersecurity best practices, draft legislation, and other regulatory governance.

Outcomes/Impacts: The research outcomes and deliverables will be a comprehensive policy guidance document and/or toolkit to assist industry participants in cybersecurity policy-making decisions. The goal of the project is to share the guidance document with as many interested individuals, departments, and agencies as possible to be used as a resource in addressing the needs identified by the Department of Homeland Security issued its Transportation Systems Sector Cybersecurity Framework Implementation Guidance. The guidance document will efficiently present the scope of existing cybersecurity governance and industry-indicated areas of need and best practices to pursue effective drafting, implementation, and enforcement of cybersecurity policy.

Lead Principal Investigator(s): Zhenkai Zhang (Clemson University)
Co-Principal Investigator(s): Long Cheng (Clemson University), Gurcan Comert (Benedict College)
Research Project Funding: Federal $95,655; Cost-share $101,128
Project Start and End Date: January 1, 2024, to December 31, 2024
Project Status: Report in Progress

Project Description: In this project, we aim to develop a secure Electronic Control Unit (ECU) hardware platform for connected vehicles utilizing the RISC-V architecture. The core innovation lies in integrating the Trusted Execution Environment (TEE) and Moving Target Defense (MTD) into the ECU. Specifically, we plan to perform the following tasks:

1. Tailoring the Keystone TEE: We will adapt the Keystone TEE specifically for ECU applications. This task involves creating a new firmware-level security monitor optimized for the CAN bus to enable device authentication and message encryption. We will also modify FreeRTOS to function as the enclave runtime, efficiently managing resources.
2. Implementing a Randomization Module: To facilitate MTD, we will incorporate a randomization module within the RISC-V core. This step will include modifying the core to include instruction set randomization logic and developing a new firmware-level configuration manager for key generation and secure storage.
3. Developing a Recovery Mechanism: A key component of our project is developing a robust recovery mechanism to ensure uninterrupted vehicle operations during an attack. This will involve setting up a fail-safe enclave that contains backup programs for each essential controller and integrating a recovery module within the configuration manager to activate these backup controllers as needed.

Moreover, we will implement the proposed platform on FPGA boards and demonstrate its effectiveness against potential attacks under the environments created in autonomous vehicle simulators. This project aims to provide a comprehensive hardware solution capable of protecting connected vehicles from a range of cyber threats, even in the presence of software vulnerabilities.

USDOT Priorities: Our project, dedicated to creating a secure RISC-V-based ECU hardware platform for connected vehicles, aligns seamlessly with the U.S. Department of Transportation's strategic priorities and RD&T goals. By fortifying ECUs against various cyber-attacks, we aim to significantly reduce the risk of incidents that can compromise passenger safety. This directly supports the DOT's priority of improving transportation safety nationwide. Moreover, our project contributes to building a more resilient transportation network. This is in line with the DOT's focus on enhancing the resilience and reliability of national infrastructure.

By tackling some of the pressing challenges in automotive cybersecurity, our project presents immediate solutions and lays the groundwork for future innovations in autonomous and connected vehicles. We envision our research findings becoming integral to the evolution of smarter, safer transportation systems, reflecting our commitment to advancing the DOT's vision of a secure and efficient transportation future.

Outputs: The key output of this project will be a RISC-V-based ECU hardware platform with rigorous security measures built in. This platform safeguards vehicles even when automotive application designers are security-oblivious or must investigate crucial security facets. Implementing the platform on FPGA boards and extensive testing in autonomous vehicle simulator environments will provide proof of concept and demonstrate the platform's effectiveness against a spectrum of cyber threats.

We anticipate several invention disclosures and potential patent filings stemming from the solutions developed in this project, particularly in hardware-level security and recovery mechanisms. Beyond the immediate TraCR center, this project also aims to foster new partnerships with automotive manufacturers, technology firms, and cybersecurity entities. These collaborations will provide practical insights and help in the real-world application and refinement of the developed solutions.

Outcomes/Impacts: Our project's outcomes are anticipated to make significant yet practical contributions to increasing the security of the transportation system. By developing a secure ECU hardware platform, we expect to enhance the safety and reliability of connected vehicles, thereby reducing the risk of cyber-attacks and associated safety hazards. This platform can be a model for future regulations or policies focusing on vehicle cybersecurity. While no patents have been filed yet, the potential for such intellectual property exists, particularly in our approaches to hardware-level security.

Implementing our research could lead to changes in industry practices, encouraging manufacturers to prioritize built-in cybersecurity measures. Ultimately, our work aims to increase the overall resilience of transportation systems while potentially reducing long-term costs associated with cyber threats and enhancing public confidence in emerging automotive technologies.

Lead Principal Investigator(s): Mizanur Rahman (The University of Alabama Tuscaloosa)
Co-Principal Investigator(s): Mashrur Ronnie Chowdhury (Clemson University), Long Cheng (Clemson University)
Research Project Funding: Federal $117,686; Cost-share $118,830
Project Start and End Date: January 1, 2024, to December 31, 2024
Project Status: Report in Progress

Project Description: Successful navigation of autonomous vehicles relies on positioning, navigation, and timing (PNT) services. Global Navigation Satellite Systems (GNSS), such as GPS (USA), BeiDou/BDS (China), Galileo (Europe), GLONASS (Russia), IRNSS/NavIC (India), and QZSS (Japan), provide PNT services. However, GNSS signals are vulnerable to unintentional interference (e.g., jamming caused by walls and ceilings in garages and tunnels, and multipath issues due to high-rise buildings in urban areas) and deliberate cyber threats (e.g., jamming and spoofing of GNSS signals). Prior research shows that the use of multi-sensor fusion systems—i.e., GNSS with inertial measurement unit (IMU) and perception sensors (PS) (e.g., camera, LiDAR, RADAR)— could complement each other and correct the individual sensor output and determine reliable navigation solution under deliberate threats and GNSS-denied environments (e.g., GNSS outage and/or INS error accumulation issue and/or PS view obstruction).

However, IMU and PS can only provide relative positioning and rely on GNSS for absolute positioning. Even advanced INS (GNSS+IMU) provide cm level accuracy; however, during GNSS outage, it could accumulate position error up to 3.80 meters in just 1 minute due to error accumulation of inertial sensors. Thus, the major research gap is to comprehensively identify and understand GNSS vulnerabilities in autonomous vehicles, investigate realistic attack modeling, detection, and develop cyber-resilient navigation solutions for GNSS-based navigation.
The overarching research goal of this project is to understand the vulnerabilities of GNSS-based navigation, develop intelligent slow-drifting cyber-attacks, develop corresponding attack detection models, and devise cyber-resilient navigation solutions to enhance the GNSS-based navigation system.

The research goal will be achieved through the following research objectives: (i) investigate and develop intelligent slow-drifting GNSS spoofing attacks by manipulating GNSS signal’s navigation data; (ii) investigate and develop GNSS cyber-attack detection algorithms for slow-drifting GNSS spoofing attacks; and (iii) develop a secure in-vehicle sensor fusion-based navigation module using deep fusion algorithms during a GNSS-denied environment. The outcomes of this project will be to implement and validate intelligent slow-drifting GNSS spoofing attack models using a GNSS receiver in both laboratory and real-world environments, evaluate GNSS cyber-attack detection algorithms against intelligent slow-drifting GNSS spoofing attacks through field testing, and demonstrate proof-of-concept of an in-vehicle sensor fusion-based cyber-resilient navigation solution in a controlled, real-world environment.

USDOT Priorities: This project is dedicated to the statutory research priority of "Reducing Transportation Cybersecurity Risks." It aligns with USDOT's strategic objectives by focusing on developing cyber-resilient navigation solutions for future transportation systems. Our goals include creating job opportunities, positioning USDOT as a global leader in the cybersecurity of transportation cyber-physical social systems, ensuring American firms lead in the global economy, and contributing to low inflation through fostering the safe, efficient, and bottleneck-free movement of goods and workers (“Economic Strength and Global Competitiveness,” “Organizational Excellence”). This project will also directly focus on the TraCR’s Research Thrust 1 “Thrust 1: Security and Resiliency.”

Outputs: This project aims to deliver the following outputs, demonstrating a proof-of-concept for:

• an intelligent slow-drifting GNSS spoofing attack model using a GNSS receiver both in laboratory and real-world environments.
• GNSS cyber-attack detection algorithms in a controlled real-world environment and
• an in-vehicle sensor fusion-based navigation solution in a controlled real-world environment.

Outcomes/Impacts: Autonomous ground vehicles promise to prevent more than 9 million accidents and 2 million injuries annually, conserve 7 billion liters of fuel and save more than 36,000 lives while reducing healthcare costs associated with accidents by $190 billion in the U.S. However, successful mission execution of autonomous ground vehicles relies on reliable localization and navigation. This research will lay the groundwork for transforming cyber-vulnerable GNSS-based navigation systems using built-in in-vehicle sensors and applying advanced deep sensor fusion techniques. Following the National Institute of Standards and Technology (NIST) cybersecurity framework, this project will be the stepping stone to develop a robust, efficient, flexible, and reliable cyber-resilient navigation platform for autonomous ground vehicles.

Lead Principal Investigator(s): Long Cheng (Clemson University)
Co-Principal Investigator(s): Zhenkai Zhang (Clemson University), Gurcan Comert (Benedict College)
Research Project Funding: Federal $139,288.50; Cost-share $139,576.50
Project Start and End Date: January 1, 2024, to December 31, 2024
Project Status: Report in Progress

Project Description: Connected and Autonomous Vehicles (CAVs) are the future of personal and public transportation. As CAVs increasingly rely on cyber-based control, navigation, and communication, security has become a pressing concern in future transportation systems. The complexity and inter-connectedness of CAVs offer myriad opportunities for security compromise, potentially resulting in unsafe operation or leakage of confidential information about the user. Zero Trust Architectures (ZTA) for networks have emerged as a fundamentally new way of approaching security. It offers new paradigms for defining and enforcing policy through various means rooted in modeling trust relationships.

The zero-trust security model does not automatically trust any user or device inside or outside the network perimeter. Instead, it enforces a set of policies (i.e., rules that are dynamically maintained and enforced) to verify and ensure the security of resources. ZTA can aid in reducing potential risks in CAVs by guaranteeing that only approved users and devices can access sensitive systems and data. This project will investigate how ZTA can be adapted to CAVs to provide fundamental protection for individual components within CAV systems and their supporting infrastructure.

USDOT Priorities: This research project aligns with USDOT’s Research Priority on Reducing Transportation Cybersecurity Risks. In May 2021, President Biden issued an executive order to enhance and improve America's cybersecurity by adopting the Zero Trust Architecture. Now, federal agencies (including USDOT) are actively working to integrate Zero Trust architecture into their existing IT environment. This project aligns with the principles outlined in this executive order, emphasizing the transition toward implementing the Zero Trust Architecture in various computer systems. The new techniques and mechanisms developed in this project will make a significant step toward implementing ZTA in CAVs.

Outputs: The research outcome includes new security methodologies, algorithms, and engineering guidelines to adopt ZTA for securing CAVs. 1) We will design and evaluate the network architecture that enforces continuous authentication and authorization to enable ZTA for CAVs. 2) We will develop a high-level policy specification language to enable ZTA for CAVs. 3) We will propose new correlation-based cyber-physical zero-trust policies to ensure the safety and resilience of autonomy-enabled vehicle systems. 4) We will design robust device fingerprinting mechanisms for continuous verification and authentication in CAVs.
We will develop and implement a prototype of the proposed ZTA-CAV systems in a realistic UAV testbed (such as Husky Vehicle).

We will conduct case studies demonstrating the performance of ZTA-CAV, such as quantifying the additional time overhead incurred by implementing our ZTA-CAV solution. We plan to collect our dataset, including data from different perception modules in CAVs, sensor data such as GPS, accelerometer, digital compass, gyroscope, and CAN bus data such as wheel speed, steering angle, pitch, and accelerometer values. We will make the testbed and evaluation results available to a broader research community to explore collaboration from academic and industry partners and the education community to engage students in learning about CAV security.

Outcomes/Impacts: The final deliverables include two or more research papers, datasets, and open-source tools. The proposed solution in this project will significantly reduce the risk of cyber-attacks and enhance the safety and security of CAVs. Dr Long Cheng is a co-PI of the project “Efficient, Cybersecure and Safe EV Operations in the Clemson Smart City Testbed,” funded by Innova EV and the South Carolina Research Authority. We will seek future collaborations with Innova EV on implementing zero trust architecture in their EVs. However, the success of this project is not dependent on other active or future projects. We will also integrate research activities into curriculum development and facilitate educational thrusts in providing research opportunities for graduate and undergraduate students, especially students from underrepresented groups.

Lead Principal Investigator(s): Jagruti Sahoo (South Carolina State University)
Co-Principal Investigator(s): Judith Mwakalonge (South Carolina State University), Nikunja Swain (South Carolina State University), Biswajit Biswal (South Carolina State University), Gurcan Comert (Benedict College)
Research Project Funding: Federal $87,291; Cost-share $87,367
Project Start and End Date: January 1, 2024, to December 31, 2024
Project Status: Report in Progress

Project Description: Connected and autonomous vehicle (CAV) technology has brought a major transformation in the transportation sector by significantly improving the mobility of people and goods through advanced communication, sensing, and computing capabilities. However, CAVs can be hacked due to vulnerabilities in the in-vehicle software, resulting in physical damage and jeopardizing the safety of drivers and passengers. By exploiting the vulnerabilities, hackers can perform malicious actions ranging from draining batteries and taking control of the steering wheel to disabling the alarm system. The existing security solutions implemented in CAVs are static and cannot withstand evolving security threats such as Advanced persistent threats (APT) and ransomware attacks. Moreover, costly update procedures leave the CAV software unpatched for a long time, making the CAVs vulnerable to new exploits.

This project aims to develop a virtualized security framework to improve the resiliency of CAV software. The framework will allow the execution of different code variants of CAV software to introduce uncertainty in the attack surface. The proposed framework will integrate the Network Functions Virtualization paradigm to implement the code variants of CAV software as virtual network functions. The proposed framework will offer the ability to optimally deploy the appropriate virtual network functions using a reinforcement learning agent [8]. The reinforcement learning agent perceives the threat environment of CAVs and provides the optimal code variant that maximizes the resiliency of CAV software while ensuring their Quality of Service (QoS) requirements.

This project aims to accomplish the following goals: 1) develop a virtualized security framework that allows fast and dynamic provisioning of different code variants of CAV software, 2) design novel and efficient algorithms designed based on game theory and Artificial Intelligence (AI) techniques including Deep Learning and Generative Adversarial Networks (GANs) to determine the optimal code variant, 3) evaluate the performance of reinforcement learning algorithm using simulations, 4) build a proof-of-concept of the proposed security framework and evaluate its performance using real-world experiments.

USDOT Priorities: This project supports the USDOT statutory research priority area “Reducing Transportation Cybersecurity Risks.” by addressing the cyber risks in CAVs that form an integral component of the transportation infrastructure. The proposed virtualized security framework provides an adaptive defense mechanism to minimize cyber-attacks. Adaptability is ensured by dynamically mutating the attack surface of CAVs according to the threat scenario and satisfying the mobility of CAVs and QoS requirements of the CAV software. This project supports the “Transformation” goal of the USDOT Strategic Plan by developing new and novel security technologies to ensure the robustness and resiliency of the transportation system.

This project will advance the state-of-the-art on security of CAV by proposing a novel security framework designed based on the network function virtualization paradigm. This project will investigate a new code diversification approach that allows the execution of different code variants of CAV software that prevents hackers from discovering the inner workings of the application code. This research shows a novel use of virtual network functions that implement the CAV software and are deployed as needed to prevent intrusions and minimize downtime. Building a robust reinforcement learning agent to automate and optimize the code diversification process is one of the important contributions of this project.

Outputs: The expected outputs of this project include a novel security framework for CAVs, a Markov Decision Process model, a new and efficient code deployment algorithm, proof-of-concept, and scientific publications.

The proposed virtualized security framework brings a new technology to dynamically mutate the CAV software and secure CAVs against intrusions, data breaches, and vehicle malfunctioning. This project will result in a Markov decision process model for representing our reinforcement learning agent using a mathematical framework. This project leads to new and efficient algorithms designed based on game theory and Artificial Intelligence (AI) techniques, including Deep Learning and Generative Adversarial Networks (GANs). These algorithms aim to maximize CAV software's resiliency in an uncertain environment characterized by dynamic strategies adopted by hackers. Proof-of-concept is one of the important outputs of this project, as it validates the suitability of the proposed security framework for practical applications. It will be demonstrated at flagship vehicular conferences such as IEEE VTC and IEEE International Conference on Intelligent Transportation Systems. Our scientific publications include peer-reviewed conferences and journal articles focusing on the proposed framework, algorithms, and evaluation using simulation and real-world experiments.

We are working to establish partnerships with agencies and/or companies interested in the research focus area: “Security and Resiliency” of the National Center for Transportation Cybersecurity and Resiliency (TraCR).

Outcomes/Impacts: The proposed security framework will improve the resiliency and robustness of the transportation system by integrating advanced cyber defense capabilities in CAVs. The proposed code diversification approach ensures the reliable and safe operation of the vehicle, even in the presence of malicious hackers. Moreover, leveraging the network functions virtualization paradigm allows fast and on-demand provisioning of an optimal code variant of CAV software that can thwart complex cyber-attacks in the transportation sector. The proof-of-concept developed during this project will showcase the potential for the practical application of the proposed framework in automotive security. Our research findings will provide stakeholders in the transportation sector, including transportation agencies and automotive manufacturers, with critical insights into the importance of an intelligent and autonomous agent that can build and provision CAV software securely.

Lead Principal Investigator(s): Mizanur Rahman (The University of Alabama Tuscaloosa)
Co-Principal Investigator(s): Mashrur Ronnie Chowdhury (Clemson University), M Sabbir Salek (Clemson University), Yingjie Lao (Clemson University), Zhenkai Zhang (Clemson University), Shaozhi Li (Clemson University)
Research Project Funding: Federal $134,847; Cost-share $137,770
Project Start and End Date: January 1, 2024, to December 31, 2024
Project Status: Report in Progress

Project Description: Blockchain-based Vehicular Ad-hoc Network (VANET) architecture has been gaining popularity due to its distributed and decentralized architecture, efficient data transmission capability, and secure data generation and broadcasting ability over VANET networks. Rating-based or trust-value-based blockchain networks can efficiently play a trusted role by setting up the proof-of-work or proof-of-stake consensus mechanisms. Such a trust management system could ensure privacy-protected and secured vehicle-to-everything communication because of its ability to ensure the veracity of the exchanged messages via a digital signature of a message sender (e.g., vehicle). However, due to the high mobility of vehicles, small key-based encryption is necessary in VANET as it requires less complex computational operations and storage.

Existing studies prove that a non-quantum computing-based or classical attack cannot generate a cyber attack on blockchain-based VANET because blockchain can identify the attacker through consensus-based or rating-based mechanisms, hashing, encryption, and its distributed nature with transparency in the public ledger-based approach. The blockchain-based architecture relies on two cryptographic mechanisms to provide security and trust: (i) check the integrity of the data itself using hash functions, and (ii) check the ownership of the data with asymmetric cryptography. However, if a quantum algorithm can break the hash function or the cryptographic algorithm, it can create security concerns for any secure communication architecture, such as blockchain, as it uses an encryption technique (mostly on subgroup-finding algorithms utilizing factorization and discrete logarithm), e.g., Rivest-Shamir-Adleman and elliptic curve digital signature algorithms. On the other hand, although prior studies have been conducted on improving the ownership mechanism of blockchain and making it quantum-safe through post-quantum cryptography and quantum key distribution, post-quantum cryptography suffers from periodicity and symmetry. It uses large-size keys, which increase the complexity of the decryption of the key, such as a lattice-based architecture.

Hash-based cryptography and multivariate cryptography exhibit a drawback in large signature sizes, leading to a larger block size and, consequently, larger memory size. Similarly, code-based cryptography encounters the issue of increasing complexity due to larger key sizes, demanding extensive memory storage, and the risk of decoding failures when utilizing smaller keys in specific scenarios. Therefore, a novel lightweight Post Quantum Cryptographic (PQC) solution, which could adapt to the dynamic VANET scenario and ensure security against quantum-based attacks, is needed according to the US NIST’s cybersecurity framework.

The overarching goal of this project is to develop a new small key-based PQC solution, the Diophantine Isogeny Key Exchange (DIKE) scheme, for VANET to ensure security against quantum-based attacks. Specifically, the objectives of this project are to (i) develop and implement a quantum-based attack model utilizing both quantum Shor’s and Grover’s algorithms on a blockchain-based VANET, which will highlight the need for a quantum-secured blockchain and (ii) formulate a new PQC solution, DIKE, which relies on the integration of Diophantine equations and isogenies to provide a secure key exchange mechanism that is resilient against quantum attacks.

USDOT Priorities: This project is dedicated to the statutory research priority of "Reducing Transportation Cybersecurity Risks." It aligns with USDOT's strategic objectives by focusing on developing a post-quantum cryptographic solution for future connected and automated transportation systems. Our goals include creating job opportunities, positioning USDOT as a global leader in the cybersecurity of transportation cyber-physical social systems, ensuring American firms lead in the global economy, and contributing to low inflation through fostering the safe, efficient, and bottleneck-free movement of goods and workers (“Economic Strength and Global Competitiveness,” “Organizational Excellence”). This project will focus on TraCR’s Research Thrust 4 “Thrust 4: Evolving Quantum Computing Threats and Opportunities.”

Outputs: This project will have the following outputs:

• a quantum-based attack model utilizing both quantum Shor’s and Grover’s algorithms and
• a new Post Quantum Cryptographic (PQC) method, DIKE, to provide a secure key exchange mechanism that is resilient against quantum attacks.

Outcomes/Impacts: The evolution of connected and automated vehicles (CAVs) to become safe, efficient, and reliable transportation components of the mainstream transportation system largely depends on innovative, rapid, and reliable technological progression. This new PQC solution, DIKE, could ensure security against quantum-based attacks in connected transportation systems, which is needed according to the National Institute of Standards and Technology (NIST)’s cybersecurity framework. Thus, this research will also directly contribute to the NIST framework.

Lead Principal Investigator(s): Shaozhi Li (Clemson University)
Co-Principal Investigator(s): Sumanta Tewari (Clemson University), Yao Wang (Clemson University), Mashrur Ronnie Chowdhury (Clemson University), M Sabbir Salek (Clemson University), Vaneet Aggarwal (Purdue University), Satish Ukkusuri (Purdue University), Gurcan Comert (Benedict College)
Research Project Funding: Federal $228,472.50; Cost-share $229,066.50
Project Start and End Date: January 1, 2024, to December 31, 2024
Project Status: Report in Progress

Project Description: In this project, we plan to develop a hybrid classical-quantum machine learning library to detect vehicle cyberattacks. By leveraging quantum supremacy, our library should improve the speed of training and the accuracy of intrusion detection systems. Specifically, we will analyze the performance of the quantum neural network in the feature extraction and the feature analysis, respectively. After understanding this performance, we will find a hybrid classical-quantum architecture that generates the best performance. In addition, we will test our hybrid library in different quantum devices, including the superconducting quantum computer and the optical quantum computer. Different quantum error mitigation techniques based on different quantum devices will be included in our library. Moreover, we will develop a tensor network approach to improve the training efficiency of the variational quantum circuits. In sum, our research focuses on investigating the architecture of the hybrid system and the optimization method in training. With our developed library, we will apply it to detect various vehicle cyberattacks, improving driving security.

USDOT Priorities: Our research objectives are developing a quantum hardware-efficient hybrid quantum-classical AI library and improving the speed of training and the accuracy of intrusion-detection systems by applying our approach. We will develop various algorithms for different quantum hardware, including superconducting qubits and trapped iron qubits. Our efficient and accurate AI library will improve the traveling public's safety, supporting the USDOT safety priority. In addition, we are using emergent quantum computers to improve the AI library, supporting the USDOT priority of embracing new technologies and fostering innovation in transportation. Moreover, our AI library enhances the resilience of transportation systems in the U. S., fitting the resilience and preparedness priority.

The classical machine learning technique has been frequently used in cybersecurity, finance, and health. However, the classical machine learning technique needs to be improved. For instance, in the training neural networks with thousands of parameters, it is easy to be stuck at a local minimal solution with zero gradients, leading to a wrong answer. In addition, the classical machine learning technique could be more efficient in training big data. Our research will address these significant problems using the quantum approach. Our work will make a significant advancement in the AI technique.

Outputs: This project will generate a new machine learning software, a hybrid classical-quantum convolutional neural network for general purposes. In addition, we will apply our software to improve the detection of the cyberattack in vehicles.

Outcomes/Impacts: The application of our developed software will increase transportation safety in the United States. Our new software will increase the accuracy of detecting cyberattacks in vehicles, preventing hackers from sending wrong signals to vehicles and causing traffic accidents. In addition, our highly accurate detection system can improve the defending system in vehicles, enhancing vehicle theft prevention.

The application of our developed software will decrease the cost of maintaining a safe driving environment. As mentioned, our hybrid machine-learning method will be more efficient than the classical machine-learning approach, which requires thousands of CPUs to perform computation. Compared to the classical computer, the quantum computer is exponentially faster than the traditional computer, significantly reducing the computational cost.