<rss version="2.0" xmlns:atom="https://www.w3.org/2005/Atom">
  <channel>
    <title>Research in Progress (RIP)</title>
    <link>https://rip.trb.org/</link>
    <atom:link href="https://rip.trb.org/Record/RSS?s=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" rel="self" type="application/rss+xml" />
    <description></description>
    <language>en-us</language>
    <copyright>Copyright © 2026. National Academy of Sciences. All rights reserved.</copyright>
    <docs>http://blogs.law.harvard.edu/tech/rss</docs>
    <managingEditor>tris-trb@nas.edu (Bill McLeod)</managingEditor>
    <webMaster>tris-trb@nas.edu (Bill McLeod)</webMaster>
    <image>
      <title>Research in Progress (RIP)</title>
      <url>https://rip.trb.org/Images/PageHeader-wTitle-RIP.jpg</url>
      <link>https://rip.trb.org/</link>
    </image>
    <item>
      <title>Algorithms to Convert Basic Safety Messages into Traffic Measures</title>
      <link>https://rip.trb.org/View/2701259</link>
      <description><![CDATA[Connected vehicles rely on short-range messaging using Basic Safety Message (BSM) data that includes information about vehicle size, speed, position, and heading (direction). In the future, all vehicles will be expected to send and receive this information to enhance safety and mobility. Exchange of BSM data among vehicles and traffic management systems will have the potential to generate traffic information that could be used to support current or develop new traffic measures such as travel time, end of queue information for work zones, road weather delay impact, and enhanced traffic signal control. This will be particularly valuable for arterial roadways and work zones in areas without instrumentation or where transportation systems management instrumentation is disrupted by construction. The aim of this research was to develop and validate algorithms that will use BSM data to estimate selected traffic measures that could be used for performance monitoring, traffic control, and traveler information.

]]></description>
      <pubDate>Tue, 12 May 2026 15:54:15 GMT</pubDate>
      <guid>https://rip.trb.org/View/2701259</guid>
    </item>
    <item>
      <title>SPR-5012: Investigating FAST Act Legislation Requirements for Bridge Load Rating Considering Emergency Vehicles</title>
      <link>https://rip.trb.org/View/2700540</link>
      <description><![CDATA[The research aim is to understand the impact of the Fixing America’s Surface Transportation (FAST) Act combined with exemptions in Indiana code on load rating of bridges and to develop recommendations that comply with legislation while reducing staff burden. The FAST Act has resulted in emergency vehicle loads that must be considered for load rating of bridges. Due to exemptions in Indiana code, this requirement applies to interstate and noninterstate bridges, leading to the posting of 1,649 bridges, mostly maintained by local agencies. The focus will be on understanding the impact on locally maintained bridges and providing guidance on compliance.]]></description>
      <pubDate>Thu, 07 May 2026 09:23:55 GMT</pubDate>
      <guid>https://rip.trb.org/View/2700540</guid>
    </item>
    <item>
      <title>TraCR Foundational Project: TraCR Collective Transportation Cybersecurity Testbeds</title>
      <link>https://rip.trb.org/View/2697454</link>
      <description><![CDATA[The National Center for Transportation Cybersecurity and Resiliency's (TraCR's)
foundational project aims to develop technological tools, prototypes, testing platforms, and facilities to ensure the cybersecurity and cyber-resilience of multimodal transportation systems and related infrastructure. The project is led by Clemson University (Clemson) under the strategic direction of Dr. Ronnie Chowdhury (Lead PI), with coordination support from Dr. Sabbir Salek (Co-PI), and involves all eight other TraCR partner institutions organized into four subgroups. A structured project governance framework, including biweekly subgroup meetings, monthly full-team coordination meetings, quarterly progress reporting and advisory board engagement, ensures alignment with project milestones, integration across teams, and effective monitoring of technical progress and deliverables. 

Clemson collaborates with Benedict College (Benedict), South Carolina State University (SCSU), and the University of Texas at Dallas (UTD) to advance a comprehensive, automated threat modeling capability for multimodal transportation systems. Building on the Transportation Cybersecurity and Resiliency Threat Modeling Framework (TraCR-TMF), the team conducts testbed-in-the-loop evaluations within Clemson’s real-world cybersecurity testbed, implementing digital-twin-based cybersecurity analysis of in-vehicle networks, and engaging state transportation agencies to assess operational transferability. Additionally, the team will work to integrate graph-based reasoning models into threat modeling, deploy supervised ModernBERT classifiers, and align with the MITRE Embedded Systems Threat Matrix to strengthen structured system-to-vulnerability mapping and improve threat coverage across transportation cyber-physical systems.

The other partner institutions will develop additional real-world and virtual testing platforms to support cybersecurity experimentation for multimodal transportation. Florida International University (FIU) and the University of Alabama at Tuscaloosa (UA) are jointly advancing the Open-Source Connected and Automated Mobility Co-Simulation (OpenCAMS) environment and related simulation platforms, integrating SUMO, CARLA, and network simulation tools, to evaluate privacy-aware multimodal large language models and post-quantum-secure C-V2X communications. Their efforts further include the development and validation of spoofing attack models targeting Basic Safety Message transmissions and multi-frequency GPS receivers, as well as investigations into backdoor-resilient perception systems and the security of vision-language models for intelligent transportation applications.

Purdue University (Purdue) and the University of California, Santa Cruz (UCSC) are advancing adversarial testing methodologies through integrated physical-virtual experimentation frameworks that combine miniature autonomous vehicle testbeds, CARLA/METS-R simulation coupling, and scenario-based vulnerability discovery. These activities include simulation-to-real validation of perception and traffic signal spoofing attacks, evaluation of V2X safety message vulnerabilities, cybersecurity analysis of shared micromobility Bluetooth pairing protocols, implementation of lightweight post-quantum cryptographic protections for vulnerable road user beacons, and closed-loop security assessments of traffic signal controller infrastructures, along with investigations of secure multimodal AI agents and memory-augmented reasoning architectures for autonomous robotic transportation systems.

In addition, Morgan State University (MSU) is enhancing its connected vehicle cybersecurity experimentation capabilities by developing replay-attack models targeting C-V2X onboard units and evaluating mitigation strategies in its real-world testbed environment, in collaboration with Clemson. These efforts quantify communication-level impacts on safety-critical applications and support the development of deployable countermeasures to strengthen resilience against wireless attack vectors affecting connected transportation infrastructure.
]]></description>
      <pubDate>Thu, 30 Apr 2026 12:19:22 GMT</pubDate>
      <guid>https://rip.trb.org/View/2697454</guid>
    </item>
    <item>
      <title>A Novel Hybrid Attack Model and A Quantum-Infused Hybrid Defense Method for Resilient Perception of Autonomous Vehicles</title>
      <link>https://rip.trb.org/View/2697419</link>
      <description><![CDATA[This project strengthens the cybersecurity and resiliency of camera-based perception for autonomous vehicles by addressing two fast-growing attack classes: universal adversarial perturbations (UAPs) and generative/deepfake-style scene manipulations that can add, alter, or remove objects in the camera feed. The team will first build and validate a novel hybrid attack that combines image-agnostic UAP noise with generative object-disappearance attacks (ODAs) using real-time inpainting to create “hallucination” driving scenes where objects are misclassified or vanish entirely. The project will also develop a quantum-enhanced hybrid defense that fuses parameterized quantum circuits with classical deepfake/manipulation detection, leveraging quantum–classical disagreement and out-of-distribution signals to robustly detect both pixel-level perturbations and semantic object edits. The project will produce deployable prototypes: (1) a real-time hybrid “malware” attack pipeline and (2) a quantum-infused hybrid detector, which will be evaluated in realistic AV scenarios and deployed for testing on connected-vehicle testbeds (e.g., Clemson University Connected Vehicle Testbed or CU-CVT and Morgan State).

]]></description>
      <pubDate>Thu, 30 Apr 2026 12:17:12 GMT</pubDate>
      <guid>https://rip.trb.org/View/2697419</guid>
    </item>
    <item>
      <title>Guiding Electronic Control Unit (ECU) Firmware Fuzzing with Hardware-Level Side-Channel </title>
      <link>https://rip.trb.org/View/2697290</link>
      <description><![CDATA[This project develops a novel electromagnetic (EM) side-channel-guided fuzzing framework for automotive Electronic Control Unit (ECU) firmware security testing. The approach addresses key challenges in ECU security research, namely that firmware is often encrypted, proprietary, and tightly coupled to hardware, making traditional instrumentation-based fuzzing impractical. By capturing and analyzing EM emanations from ECUs during execution, the framework estimates code coverage without requiring firmware modification, instrumentation, or rehosting. The system integrates this EM-based coverage feedback into a fuzzer to guide test case generation via Controller Area Network (CAN) bus communication. The project will conduct extensive fuzzing campaigns on real automotive ECUs from various manufacturers to discover zero-day vulnerabilities and enhance vehicle cybersecurity. ]]></description>
      <pubDate>Wed, 29 Apr 2026 16:47:58 GMT</pubDate>
      <guid>https://rip.trb.org/View/2697290</guid>
    </item>
    <item>
      <title>Prototype Development and Pilot Deployment of Ground-Based Intelligent Infrastructure for Resilient Positioning, Navigation, and Timing</title>
      <link>https://rip.trb.org/View/2696990</link>
      <description><![CDATA[Global Navigation Satellite Systems (GNSS), such as the Global Positioning System (GPS), form the backbone of modern positioning, navigation, and timing (PNT) services. However, these space-based systems are inherently vulnerable to cyberattacks such as jamming, spoofing, as well as unintentional interference, including signal blockage, particularly in dense urban areas, indoor environments, and adversarial environments. The growing dependence on GNSS, driven by the rapid adoption of autonomous and connected systems, has exposed a single point of failure in the global PNT infrastructure. GPS signals are extremely weak at the Earth’s surface, enabling low-cost jammers or spoofers to easily disrupt receivers. In response to the 2020 Executive Order on strengthening national resilience through responsible use of PNT services signed by President Donald J. Trump, US DOT, the Department of War (DoW), and the Department of Homeland Security (DHS) have jointly emphasized the need for complementary and backup PNT capabilities that are interoperable and independently capable of sustaining precision timing and navigation for critical infrastructure during GNSS outages or cyberattacks. The research goal is to develop and demonstrate a prototype ground-based, GPS-compatible, cyber-secure PNT architecture that can generate, synchronize, and broadcast authenticatable GPS-like signals from a network of ground-based nodes, allowing existing GPS receivers to obtain valid PNT solutions without hardware modification. This goal will be achieved through the following specific research objectives: (1) Design and generate authenticable GPS-compatible terrestrial signals that replicate the L1 C/A (coarse/acquisition) waveform while embedding virtual ephemeris and adjusted clock-offset parameters to enable accurate and PNT computation from ground transmitters. (2) Develop intelligent terrestrial nodes (at least four nodes) equipped with chip-scale atomic clocks, edge computer, and transmitters to establish a distributed ground-based PNT architecture. (3) Synchronize terrestrial nodes with a master clock using precision timing distribution techniques to maintain consistent and reliable time alignment across the network. Real-Time Kinematic (RTK) positioning and differential methods will also be explored using the GEODNET hub within the UA network. (4) Demonstrate that an off-the-shelf GPS receiver can deliver a valid PNT solution using terrestrial signals through software-only modifications, thereby validating the practicality, backward compatibility, and deployment readiness of the proposed system.
]]></description>
      <pubDate>Wed, 29 Apr 2026 16:45:17 GMT</pubDate>
      <guid>https://rip.trb.org/View/2696990</guid>
    </item>
    <item>
      <title>Artificial Intelligence (AI)-Enabled Post-Quantum Cryptography for Real-World Deployment of Secure and Resilient Communication for Intelligent Transportation Systems</title>
      <link>https://rip.trb.org/View/2696971</link>
      <description><![CDATA[Cellular Vehicle-to-Everything (C-V2X) communication, standardized in 3GPP Release 14/15 PC5 sidelink mode, is the US DOT-approved technology for direct V2V (vehicle-to vehicle)/ V2I (vehicle-to-infrastructure) communications in the 5.9 GHz band. Current standards and specifications (e.g., SAE J3161 and USDOT/ITE RSU requirements) mandate PC5 Mode 4 operation to enable interoperable safety messaging using conventional cryptographic methods, such as Elliptic Curve Cryptography (ECC). However, existing cryptographic methods are vulnerable to quantum-computing-based attacks. Thus, integrating Post-Quantum Cryptography (PQC) into C-V2X communication is imperative to ensure future resilience. However, National Institute of Standards and Technology (NIST)-standardized PQC algorithms introduce large key sizes and computational complexity, resulting in significant latency and bandwidth overhead. These effects risk violating the 100-ms end-to-end delay requirement for 10 Hz Basic Safety Messages (BSMs) and can congest the 5.9-GHz safety channel. Moreover, the direct integration of PQC into current communication standards, such as IEEE 1609.2 and ETSI, poses challenges because these frameworks were originally designed for lightweight ECC-based operations. 
Similarly, post-quantum Homomorphic Encryption (HE) offers robust privacy protection by allowing computation directly on encrypted data without decryption; however, its high computational cost and ciphertext expansion currently limit its use in latency-critical V2X and infrastructure-to-infrastructure (I2I) scenarios. Therefore, deploying PQC and HE within operational testbeds demands optimized scheduling, resource allocation, and adaptive algorithm management to balance cryptographic strength with real-time constraints. To address these challenges, this project aims to develop and evaluate artificial intelligence (AI)-enabled PQC through real-world prototype implementation and testbed integration, thereby enabling the real-world deployment of secure and resilient communication in intelligent transportation systems. Specifically, the objectives of this project are: (i) implementation and real-world evaluation of an AI-enabled PQC integration and dynamic switching framework for C-V2X communication; (ii) real-world evaluation of a privacy-preserving roadside unit (RSU)-Cloud (I2C) communication pipeline using post-quantum homomorphic encryption; and (iii) development of a federated learning framework for collaborative PQC selection policies. To address the USDOT and TraCR 2025–2026 priorities, this project emphasizes field-tested prototypes and operational validation, rather than simulation-only evaluation, to ensure deployment relevance. This project will directly contribute to the deployment of PQC-enabled V2X communication for a secure and reliable connected transportation system.
]]></description>
      <pubDate>Wed, 29 Apr 2026 16:39:01 GMT</pubDate>
      <guid>https://rip.trb.org/View/2696971</guid>
    </item>
    <item>
      <title>Resilient Software-Defined Vehicle Platform Architectures with Secure Live Migration</title>
      <link>https://rip.trb.org/View/2696966</link>
      <description><![CDATA[Modern vehicles use software-defined vehicle (SDV) platforms that integrate functions via virtualization, but current designs lack resiliency against security incidents or hardware obsolescence. This project aims to enhance vehicle security by developing and evaluating secure live migration techniques for virtual machine (VM)-based workloads. By allowing actively running services to move between electronic control units (ECUs) without interruption, the project enables real-time upgrades and mitigation of cyberattacks within next-generation zonal architectures.

]]></description>
      <pubDate>Wed, 29 Apr 2026 16:36:12 GMT</pubDate>
      <guid>https://rip.trb.org/View/2696966</guid>
    </item>
    <item>
      <title>Investigating Driver Behavior Under Cyberattacks in Connected Vehicle Environments: Phase II</title>
      <link>https://rip.trb.org/View/2696964</link>
      <description><![CDATA[Phase II will examine driver behavior and decision-making under cyberattacks in connected-vehicle contexts using high-fidelity, human-in-the-loop driving simulators at Morgan State University (urban) and Clemson University (suburban). The team will develop reusable Cyberattack Injection Modules (CIMs) for UCWinRoads and SimCreator and validate one vehicle-centered attack (false blind spot warning) and two infrastructure-centered attacks (falsified signal phase-and-timing and “phantom” signal-ahead information). The study will leverage IRB-approved human-subject testing, conduct a Maryland MVA pilot demonstration, and curate a Standardized Attack Scenario Library (ASL) for replication and training use.]]></description>
      <pubDate>Wed, 29 Apr 2026 16:34:42 GMT</pubDate>
      <guid>https://rip.trb.org/View/2696964</guid>
    </item>
    <item>
      <title>Cyber-CAT: Prototyping and Experimental Demonstration of Cyberattack Mitigation in Connected and Automated Traffic (CAT)</title>
      <link>https://rip.trb.org/View/2696946</link>
      <description><![CDATA[The need for research on experimental prototyping and demonstration of cyberattack mitigation strategies in Connected and Automated Traffic (CAT) is critical as modern transportation systems increasingly depend on interconnected and automated vehicle operations. With the growing adoption of Vehicle-to-Everything (V2X) communication, the interactions and cascading effects that arise within real connected traffic systems caused by cyberattacks need to be investigated in a timely manner. 
The proposed project will deliver an integrated experimental framework and multi-layer mitigation strategy to strengthen cybersecurity in CAT systems. Key highlights include: Cyber-CAT System Development: Establish a real and virtual hybrid testbed integrating CAVs, infrastructure, and C-V2X communication to enable realistic testing of cyber threats and countermeasures; Multi-layer Mitigation: Prototype and validate Zero-Trust-LLM-assisted mitigation and artificial intelligence (AI)-based intention sharing mitigation to maintain safety under attack conditions; and Guidance and Knowledge Transfer: Produce data and design guidelines to inform US DOT and industry on C-V2X security standards and resilient corridor deployment.
]]></description>
      <pubDate>Wed, 29 Apr 2026 12:27:54 GMT</pubDate>
      <guid>https://rip.trb.org/View/2696946</guid>
    </item>
    <item>
      <title>SentinelLab: A Plug-in Online Defender Testbed for Connected and Autonomous Vehicle (CAV) </title>
      <link>https://rip.trb.org/View/2696945</link>
      <description><![CDATA[This project develops SentinelLab, a closed-loop defender testbed designed to transform how Connected and Automated Vehicle (CAV) cybersecurity is validated. Currently, most research stops at detecting anomalies; this project bridges the gap to active defense by integrating the METS-R traffic simulator with the CARLA photo-realistic sensor simulator. The testbed utilizes a “Recognize, Decide, Act” framework. The project employs a multimodal Large Language Model (LLM) to recognize specific attack families (e.g., message replay, breaking provocation) from noisy streaming signals. It then uses an online Defender Workbench to decide on mitigation strategies via plug-in policies and automatically executes these actions in the simulation. This system enables researchers and public agencies to prepare defenses against realistic threats and quantify their impacts on safety and mobility.]]></description>
      <pubDate>Wed, 29 Apr 2026 11:32:19 GMT</pubDate>
      <guid>https://rip.trb.org/View/2696945</guid>
    </item>
    <item>
      <title>Game-Theoretical Approach for Cyberattack Modeling and Deep Learning-Based Resilience of Connected Automated Vehicles</title>
      <link>https://rip.trb.org/View/2696944</link>
      <description><![CDATA[The current state of practice in security research on connected automated vehicles (CAVs) does not consider how adversaries may evolve over time and adapt to defense strategies. This proposed research takes the state of practice well beyond the current focus on anomaly detection towards strategic response in defense strategies by developing attack defender models with game-theoretical models. Game theory provides a framework to study the strategic interactions between defenders and adversaries with conflicting objectives. Given the above background, this study will design strategic games to study attacker and defender strategies for cyber deception, as well as algorithms to compute equilibrium or optimal defense strategies in a CAV environment. Real-world data from CAV experiments conducted by PIs will be used to design game theory models in a CAV environment. The study will design two strategic games, namely a zero-sum game and a Stackelberg security game, to formalize the interactions between attackers and defenders by devising a strategic comparison between a zero-sum game and a Stackelberg security game. The proposed game models define payoff functions that capture the trade-offs between model accuracy and the success rates of attacker and defender. The dynamic attacker-defender strategies mimic real-world applications and provide the ability to provide alerts to traffic management center operators for performing cyber incident response in a timely manner, which has attracted the Virginia Department of Transportation’s (VDOT’s) interest. VDOT will serve as a partner to help with real-world implementation. ]]></description>
      <pubDate>Wed, 29 Apr 2026 11:17:14 GMT</pubDate>
      <guid>https://rip.trb.org/View/2696944</guid>
    </item>
    <item>
      <title>Cyberattack Resilience in Cooperative Driving Automation Using Experimental Data and Federated Agents: Phase II</title>
      <link>https://rip.trb.org/View/2696905</link>
      <description><![CDATA[Cooperative driving automation or Connected and Automated Vehicles (CAVs) are rapidly taking over modern intelligent transportation systems. The proliferation of CAVs has also intensified concerns around cybersecurity and data privacy. The added communication involved in these driving maneuvers serves as a vulnerable attack surface. The data communicated in Basic Safety Messages (BSMs) of CAVs is highly safety-critical, thus requires secure processing and sharing. Traditional security strategies are mostly machine learning-based that rely on centralized data processing and storage. The centralized servers act as single-point-of-trust, which is vulnerable to failure, compromising data privacy, and adding overhead to communication. To address these challenges, Federated Learning (FL) has emerged as a distributed learning paradigm that enables CAV agents to locally train models and only share model parameters with a global server for updates. This eliminates the need for raw data sharing, which preserves the privacy of sensitive data transfer during CAV communication and reduces the risk of single-point failure.
Despite the benefits of FL, it is still susceptible to threats like poisoning attacks, inference-based adversaries and model manipulation. The model parameters are not secured while shared iteratively between local and global agents. It is possible for adversaries to deliberately inject anomalies into the local model updates, thereby degrading the accuracy of the global model or compromise the individual local agents. To mitigate this inherent problem of FL, Blockchain serves as the apt solution. Blockchain technology is lightweight, fully decentralized data storage framework that replaces conventional centralized databases by providing immutability and tamper-proofing to the stored data. The Secure Hashing Algorithm (SHA) and smart contracts employed by Blockchains facilitate trust and accountability in this storage solution. This research will integrate blockchain with FL to secure the training data shared between FL’s distributed agents. Due to the distributed nature of both frameworks, they complement each other well and are completely compatible for integration. 
]]></description>
      <pubDate>Tue, 28 Apr 2026 16:15:44 GMT</pubDate>
      <guid>https://rip.trb.org/View/2696905</guid>
    </item>
    <item>
      <title>Compositional Modeling and Attack Analysis of End-to-end Autonomous Vehicles</title>
      <link>https://rip.trb.org/View/2696904</link>
      <description><![CDATA[The architecture of autonomous driving systems is shifting from modular pipelines to end-to-end systems powered by advanced Artificial Intelligence (AI) and Vision-Language Models (VLMs). This project develops a comprehensive framework for the compositional modeling, security analysis, and resilience testing of these next-generation systems. The research team will create formal models to abstract the behavior of AI components and build an AI-powered vulnerability analysis engine to identify semantic attacks. The project culminates in a high-fidelity, open-source software testbed that integrates these models to simulate attacks and evaluate the resilience of autonomous vehicles and drones.]]></description>
      <pubDate>Tue, 28 Apr 2026 15:45:25 GMT</pubDate>
      <guid>https://rip.trb.org/View/2696904</guid>
    </item>
    <item>
      <title>Secure Multi-Modal Transportation Artificial Intelligence (AI) at Run-Time</title>
      <link>https://rip.trb.org/View/2689389</link>
      <description><![CDATA[This research develops robust and secure artificial intelligence (AI) systems for smart transportation that could defend against novel adversarial attacks with high performance. The project addresses critical vulnerabilities in Test-Time Adaptation (TTA) mechanisms and Multimodal Large Language Models deployed in transportation systems. Through new attack discovery, effective defense framework development, and deployable prototype systems, this work ensures that AI technologies can be safely deployed in safety-critical transportation applications. The research delivers practical solutions including natural scene adversarial attack frameworks, sharpness-aware minimization based TTA defenses, event-conditioned representation compression for efficient multimodal AI, and adversarially robust multimodal fusion architectures.]]></description>
      <pubDate>Tue, 28 Apr 2026 15:43:28 GMT</pubDate>
      <guid>https://rip.trb.org/View/2689389</guid>
    </item>
  </channel>
</rss>