C²M² would like to thank Dr. Zadid Khan, Walmart for taking part in our C²M² CPS Frontiers Series. Dr. Khan spoke on April 21, 2022.

Seminar Title

Cybersecurity of Connected Automated Vehicles in Transportation Cyber-physical Systems with Artificial Intelligence.

Seminar Abstract

The transportation system in the US is transforming into intelligent cyber-physical systems (CPS) with the advancement and merging of sensing, computer, and communication technologies. One of the core components of the transportation CPS (TCPS) is connected and automated vehicles (CAVs). Improving the cybersecurity of in-vehicle networks and automated vehicle applications are critical areas that need further attention. In this talk, Dr. Zadid Khan demonstrates novel methods and models to improve the cybersecurity of CAVs using artificial intelligence under different cyberattack scenarios. This talk is composed of two primary topics. In the first topic, an anomaly detection model is developed for a vehicle’s in-vehicle controller area network (CAN). The model is evaluated using CAN datasets from two real vehicles. Test results show improvement in detection accuracy over baseline models with the model. The second topic focuses on the cybersecurity of automated vehicles. Under this topic, a hybrid defense method is developed that protects an autonomous vehicle’s deep learning models against adversarial attacks. These deep learning models are used for traffic sign classification. This method uses a combination of random filtering, ensembling, and local feature mapping methods to improve the resilience of the traffic sign classifier used by autonomous vehicles. Analysis shows that this defense method improves baseline defense strategies in making an autonomous vehicle sign classifier resilient against different types of adversarial attacks.

Speaker Bio

Dr. Zadid Khan is a Senior Data Analyst in the supply chain (transportation) department at Walmart, Inc. He received his Ph.D. and M.Sc. in Civil Engineering (transportation major) from Clemson University and B.Sc. degree in Electrical Engineering from the Bangladesh University of Engineering and Technology (BUET). His primary research focus is cybersecurity and reliability of transportation cyber-physical systems (TCPS) and connected automated vehicles (CAVs). Within the TCPS and CAV domains, his research interests are data science, machine/ deep learning, computer networking, data analytics, cloud computing, and optimization. During his Ph.D. and M.Sc., Dr. Zadid Khan worked under the supervision of Dr. Mashrur “Ronnie” Chowdhury on multiple C²M² supported research projects.  

C²M² would like to thank Manveen Kaur, Clemson University for taking part in our C²M² CPS Frontiers Series. Manveen Kaur spoke on April 28, 2022.

Seminar Title

The Design and Validation of an ICN-Enabled Hybrid Unmanned Aerial System.

Seminar Abstract

This talk will present a measurement study that evaluates a novel Information-Centric Networking (ICN)- enabled Hybrid Unmanned Aerial Vehicle (UAV) System called IH-UAS. IH-UAS leverages ICN and an innovative system model integrating broker-based publish-subscribe message dissemination with a decentralized architecture to form an ad hoc (infrastructure-less) UAS to carry out military missions. The goal of this study is to design a system that pushes decision-making to the UAV swarm on the battlefield such that mission tasks are completed more reliably and in less time than traditional centralized UAV-based missions. We use theoretical and measurement-based analysis to validate the system. Through experiments conducted using a simplified variant of a Coordinated Search and Tracking (CSAT) application in IH-UAS, we demonstrate that IH-UAS performs better than the same application operating in a traditional centralized solution. We also discuss how the broker placement and the number of brokers are critical to application performance.

Speaker Bio

Manveen Kaur is a Ph.D. student in the School of Computing at Clemson University. Her research focuses broadly on wireless networks and systems, specifically designing, implementing and evaluating systems design and communication methods for emergent Internet of Things (IoT) systems. The two primary systems used in her research are Connected Vehicular Networks and Unmanned Aerial Vehicle (UAV) swarms. She is currently working on a solution that provides efficient system connectivity and data dissemination services for resource-constrained IoT systems running data compute-intensive applications with strict performance requirements. Manveen obtained her M.S. degree in Electrical and Computer Engineering from The Ohio State University. She worked in the media-streaming industry as a Systems Engineer before joining Clemson University.  

Applications using dedicated short-range communication (DSRC) are being developed to prevent automobile accidents. Many DSRC implementations, applications, and network stacks are not mature. They have not been adequately tested and verified. This study illustrates security evaluation of a DSRC wireless application in vehicular environments (DSRC/WAVE) protocol implementation. We set up a simulation of a working road side unit (RSU) on real DSRC devices. Our experiments work on the Cohda testbed with DSRC application wsm-channel. We extended the functionality of wsm-channel, an implementation of WAVE short message protocol (WSMP) for broadcasting GPS data in vehicular communications, to broadcast car information, and RSU instructions. Next, we performed Denial of Service attacks to determine how few packets need to be dropped to cause automobile crashes. Hidden Markov Models (HMM) are constructed using sniffed side channel information since operational packets would be encrypted. The inferred HMM tracks the protocol status over time. To test HMM’s ability to predict which packets will be dropped, we used a simulation-based experiment and implemented a DSRC-supported stop light application. Using these simulations, we were able to show that we could accurately identify the packets we needed to drop by using timing and packet size side channels. The attack simulation using inter-packet delay side-channel features worked best to drop necessary packets with a 2.5 % false positive rate (FPR) while the attack using a packet size side-channel worked with a 9.5% FPR.

• Final Report
• Technology Transfer Report
• Report Data

C²M² invites you to join us in welcoming Anton Bezuglov, Vericast as a part of our C²M² Distinguished Speaker Series. 

Seminar Title

Tuning Personalized Recommendations with the Multi-Armed Bandit Approach

Seminar Abstract

While content-based and collaborative filtering are well-known types of product recommendation engines, their real-life application is sometimes problematic. This is due to the cold start issues, data veracity, or the multitude of hyperparameters. A combination of these factors makes it difficult to forecast which approach will produce the best fit products and for how long it will remain the best. This talk focuses on the optimization scheme where multiple algorithms run simultaneously and compete for the user audience. This optimization is more flexible when compared to A/B or A/B/n testing and it saves more user conversions.

Speaker Bio

Anton Bezuglov has a Ph.D. in Computer Science and Engineering from the University of South Carolina. He worked in academia for 12 years — first as a  Professor of Computer Science (Benedict College), then as a Professor of Data Analytics (Buena Vista University). Since 2020, he has been a Data Scientist at Vericast. His areas of focus are digital advertising, ad optimization, brand safety, and personalized recommendations.

C²M² would like to thank Dr. Mhafuzul Islam, General Motors R&D for taking part in our C²M² CPS Frontiers Series. Dr. Islam spoke on March 18, 2022.

Seminar Title

Hybrid Quantum-Classical Machine Learning for Cloud-supported In-Vehicle Cyberattack Detection

Seminar Abstract

Quantum computing is an emerging area, where using the computational power of a quantum computer a vast number of calculations can be performed simultaneously compared to classical computers. Though utilizing the true power of a quantum computer is far-fetched, a hybrid setup of quantum and classical computers can achieve a computational advantage over a classical computer only. In this talk, Dr. Mhafuzul Islam will present one of his studies of using a hybrid quantum-classical neural network to detect a cyberattack on the in-vehicle communication network. This study found that using a hybrid quantum-classical neural network, it is possible to achieve a higher attack detection accuracy compared to a quantum-only neural network, and a classical neural network.

Speaker Bio

Dr. Mhafuzul Islam is currently working as a Senior Researcher in General Motors R&D, USA. He received his Ph.D. and M.Sc. in Civil Engineering from Clemson University and B.Sc. degree in Computer Science and Engineering from the Bangladesh University of Engineering and Technology (BUET). His research area includes transportation cyber-physical systems with an emphasis on data-driven connected autonomous vehicles. During his Ph.D. and M.Sc., Dr. Mhafuzul Islam worked under the supervision of Dr. Mashrur Chowdhury and worked on multiple C2M2 funded research projects. He previously worked as a senior software engineer at Infineon Technologies AG, Germany.  

C²M² invites you to join us in welcoming Elham “Ellie” Masoomkhah, Harvard University, as a part of our C²M² Distinguished Speaker Series. 

Seminar Title

Racist Highways!

Seminar Abstract

US highway system is considered one of the critical infrastructures in the history of transportation in this country. We all use them! We all enjoy the goods and services delivered to us through them. Our cities and towns are interconnected by them. It is almost impossible to imagine our lives without the vast network of interstates that spatially connect us to one another in today’s world. However, there were and still are social and economic costs to these enormous networks of roads. This talk will discuss some of those spatial disparities that are disproportionately burdened by minorities.

Speaker Bio

Ellie is a Geography Information Systems (GIS) enthusiast and actively working to increase spatial awareness by learning, using, teaching, and sharing GIS. She started her GIS career in 2014 at Clemson University Campus Planning and Design as a GIS Specialist. From there, she joined Clemson Center for Geospatial Technologies as GIS Manager, where she managed the daily operations of the Geospatial Center (CCGT). This included managing the geospatial technology consultation services, the implementation, and management of cloud-computing and web-based GIS platforms, sustaining and strengthening the GIS program as the primary provider of GIS-based services, research consultation, and user education to Clemson University students, faculty, and staff from all departments across campus and developing customized training materials for class instruction and project support.

She considers her work as part of CCGT as a once-in-a-lifetime opportunity that she was fortunate to have as she participated in the development and growth of one of the best geospatial centers nationwide. She is now with the Center as a GIS Consultant since she left Clemson in September 2021 to work as GIS Manager at Harvard University Planning and Design office.

Ellie earned a B.Sc. in Architectural Engineering from the National University of Iran in 2012. She worked as an architect and an urban planner in Iran for a while before leaving home. She obtained her Master’s degree in City and Regional Planning program at Clemson University.

C²M² would like to thank Dr. Taimi Olsen, Clemson University, for taking part in our C²M² Distinguished Speaker Series on January 26th, 2022.

Seminar Title

Inclusive Teaching for Today’s Engineering and Science Students

Seminar Abstract

This interactive talk will address our understanding of and interactions with today’s students, as we think through mindsets, motivations, and engagement practices that can work with Gen Zers and be inclusive and support all students. As experienced educators, how do we know what practices are inclusive, and how do we avoid unintentionally adding barriers to student academic success in higher ed?

Speaker Bio

Dr. Taimi Olsen is an educator and faculty developer with four decades of experience in higher education. She is devoted to universal, transparent teaching, bringing her passion to teaching first generation students, women and students of color, international students, and students with disabilities—and to bring educational support to faculty in these positionalities.

C²M² is happy to share our latest collaborative effort with Clemson’s College of Engineering, Computing and Applied Science to showcase our research in the emerging field of Quantum Artificial Intelligence for Transportation Cyberphisycal Systems. This video features Dr. Mashrur “Ronnie” Chowdhury and explains this exciting new field of research.

Attribution theory refers to the psychological phenomenon where one person tries to perceive others’ cognitive behavior by ascribing their own emotions, opinions, and desires. For instance, while passing at an intersection, a driver expects that the maneuver of other drivers coming from the opposite direction or conflicting movements would be like their own. When expected behaviors do not match the opposite or conflicting movement driver’s future behaviors, a collision is likely to occur. This research investigated the application of attribution theory to assume the opposing drivers’ cognitive behavior and performance at a highway intersection. This phenomenon was evaluated by utilizing the second Strategic Highway Research Program (SHRP-2) and National Highway Traffic Safety Administration (NHTSA) data sources. From the data analysis, it was observed that drivers aged 25-34 years were involved in the highest number of fatal accidents from 2009-2018 in the USA. Besides, it was found that younger drivers (aged 20-years old or less) contributed fewer fatal collisions (44,404 crashes) than elderly drivers (aged 65-years old or more, 62,572 crashes). The impact of the attribution theory and driver age in highway intersection-related collisions were examined from simulation models. From simulations, it was observed that there was a high possibility of collisions when an elderly driver was turning left. In this research, the combination of an elderly driver turning left, and the younger driver going straight resulted in the highest number of collisions compared to other groups. The key findings confirm elderly and younger drivers have different driving behavior that could be ascribed to their attribution. These results can assist transportation agencies to develop training and design strategies to better accommodate elderly drivers due to their declined physical and cognitive abilities and improve drivers’ education programs for younger drivers.

• Final Report
• Technology Transfer Report
• Report Data

C²M² invites you to join us in welcoming Dr. Yunyi Jia, Clemson University International Center for Automotive Research, as a part of our C²M² Distinguished Speaker Series.

Seminar Title

Cloud-based Collaborative Road Condition Monitoring using In-Vehicle Smartphones

Seminar Abstract

Ensuring the safety of transportation systems requires monitoring the conditions of roads. Traditional monitoring and inspection of road conditions require surveyors to go along the roads to search for defects. Such processes are very time consuming, expensive, and labor intensive. Recent automated road condition monitoring approaches require specially equipped vehicles with specific sensors associated with complex processing systems, which is still time consuming and expensive. In addition, all these approaches cannot provide real-time monitoring of road conditions. Therefore, we investigate a more efficient and cost-effective approach to monitor the road conditions through cloud-based collaborative monitoring using in-vehicle smartphone data from conventional vehicle users. We first collect motion and vision data from smartphones and leverage deep learning approaches to detect road conditions. We further send the information from various smartphones to the cloud and fuse the detection results in the cloud in order to generate more complete and precise detection results. Based on the fusion, the road conditions are updated in real time from data reported by the smartphone users and visually displayed in a map which can be viewed by the concerned authorities through a webpage.

The project is sponsored by the USDOT Center for Connected Multimodal Mobility (C²M²) and is a collaboration between Benedict College and Clemson University.

Speaker Bio

Dr. Yunyi Jia is the McQueen Quattlebaum assistant professor in the Department of Automotive Engineering at Clemson University. He directs the Collaborative Robotics and Automation (CRA) Lab and his research focuses on robotics, autonomous vehicles, and advanced sensing systems. He has been the receipt of the SAE Ralph R. Teetor Educational Award, NSF CAREER Award, NSF CPS CRII Award, and SAE Trevor O. Jones Outstanding Paper Award. He received his Ph.D. in Electrical Engineering from Michigan State University in 2014. He is currently a senior member of IEEE and a member of ASME and SAE.