Advertise on podcast: Trusted CI podcast

While AI is increasingly present in clinical practice especially for medical imaging, it is imminent to ensure cybersecurity of imaging diagnostic AI models. Newly advanced adversarial attacks pose a threat to the safety of medical AI models, but little is known about the characteristics of this threat. Medical adversarial attacks may lead to serious consequences including patient harm, liability of healthcare providers, and other ethical issues or crimes. It is imperative to study this cybersecurity issue to mitigate potential negative consequences and to ensure safety of health care. In this talk, the speaker will discuss cyber vulnerabilities of deep learning-based medical imaging diagnosis models under adversarial attacks, show real-world experiments on how adversarial attacks can fool AI models to decrease diagnosis performance and to confuse experienced radiologists, and present several methods of defending adversarial attacks to secure AI models in medical imaging applications. Speaker Bio: Shandong Wu, PhD, is a Professor in Radiology, Biomedical Informatics, Bioengineering, and Intelligent Systems at the University of Pittsburgh. Dr. Wu leads the Intelligent Computing for Clinical Imaging (ICCI) lab, and he is the founding director of the Pittsburgh Center for AI Innovation in Medical Imaging. Dr. Wu’s work focuses on developing trustworthy medical imaging AI for clinical/translational applications. Dr. Wu's lab received multiple research awards such as the RSNA Trainee Research Award twice in 2017 and 2019, the 2021 AANS Natus Resident/Fellow Award for Traumatic Brain Injury, the 2025 SPIE Imaging Informatics Best Paper Award, etc. Dr. Wu’s research is supported by NIH, NSF, multiple research foundations, Amazon AWS, Nvidia, and many institutional funding sources. Dr. Wu has published > 190 journal papers and conference papers/abstracts in both the computing and clinical fields. His research has been featured in hundreds of scientific news reports and media outlets in the world.

The challenge of providing end to end trust and security for operational technology systems has been a growing challenge and increasingly imperative. An IEEE effort was begun in 2016 to tackle that challenge, resulting in the publication of the first IEEE/UL TIPPSS standard (IEEE/UL 2933-2024) and the awarding of the 2024 IEEE Standards Association Emerging Technology Award to the TIPPSS standard working group. The goal of the TIPPSS standard, which is envisioned to be a family of standards, is to improve Trust, Identity, Privacy, Protection, Safety and Security (TIPPSS) for cyber-physical systems, beginning with Clinical Internet of Things and expanding to research infrastructure, the energy grid, distributed energy resources, and more. In this webinar we will discuss the initial IEEE/UL TIPPSS standard for clinical IoT data and device interoperability, the details of the technical and process elements of the standard, and the opportunity to apply it to all operational technology. Future TIPPSS presentations planned include "TIPPSS for navigating a changing cybersecurity landscape at the Electron-Ion Collider and other scientific research facilities" in collaboration with Brookhaven National Laboratory at the ICALEPS 2025 conference (The 20th International Conference on Accelerator and Large Experimental Physics Control Systems) September 20-26, 2025 in Chicago, bringing the TIPPSS discussion to research infrastructure and the IT systems that support it. Trusted CI's initiatives in Secure by Design and the Trusted CI Operational Technology Procurement Vendor Matrix are very complementary to the TIPPSS initiative, and there is more we can do as a community in this effort together. Join us to discuss the imperatives and possibilities. Speaker Bio: Florence Hudson is Executive Director of the Northeast Big Data Innovation Hub at Columbia University, leading over $10M in projects funded by the National Science Foundation, National Institutes of Health, and Department of Transportation. She is also Founder & CEO of FDHint, LLC, a global advanced technology consulting firm. A former IBM Vice President and Chief Technology Officer, Internet2 Senior Vice President & Chief Innovation Officer, Special Advisor for the NSF Cybersecurity Center of Excellence, and aerospace engineer at the NASA Jet Propulsion Lab and Grumman Aerospace Corporation, she is an Editor in Chief and Author for Springer, Elsevier, Wiley, IEEE, and other publications. She leads the development of global IEEE/UL standards to increase Trust, Identity, Privacy, Protection, Safety and Security (TIPPSS) for connected healthcare data and devices and cyberphysical systems, and is Vice Chair of the IEEE Engineering Medicine & Biology Society Standards Committee. She earned her Mechanical and Aerospace Engineering degree from Princeton University, and executive education certificates from Harvard Business School and Columbia University.

The democratization of high-performance computing (HPC)—driven by a paradigm shift toward cloud-based solutions—has unlocked unprecedented scalability in data sharing, interdisciplinary collaboration, and large-scale analytics. Yet, despite these advancements, the lack of strong privacy protection mechanisms, particularly for sensitive or regulated data, remains a significant barrier preventing critical domains from fully leveraging cloud HPC. In this webinar, I will present our group’s research toward enabling a practical confidential HPC paradigm—one that empowers HPC providers to securely process sensitive workloads with provable security and privacy guarantees. My talk will be structured around three key pillars that underpin our approach: practical data-in-use security, data governance and compliance, and usability. First, I will introduce our vision for a next-generation trusted execution environment (TEE) architecture tailored for HPC—designed to deliver HPC-grade efficiency for large-scale, parallel workloads, while upholding strict data-in-use security guarantees. Second, I will discuss how we leverage formal methods to validate compliance with complex governance and data-sharing policies—ensuring that even dynamic, multi-party workloads can remain policy-aligned. Finally, I will share our ongoing work in developing new usability frameworks and programming abstractions designed to make confidential computing accessible to domain scientists—lowering the barrier for adoption without requiring expertise in cryptography or secure systems. Speaker Bio: Chenghong Wang is an Assistant Professor in the Luddy School of Informatics, Computing, and Engineering at Indiana University Bloomington. He is a core faculty member of the Security and Privacy in Informatics, Computing, and Engineering (SPICE) group and affiliated with the NSF Center for Distributed Confidential Computing (CDCC). His research focuses on building full-stack solutions for privacy-preserving data sharing and analytics (PPDSA), bridging theory, systems, and architectural design. His work spans trusted execution environments, differential privacy, applied cryptography, and secure data systems. Dr. Wang's research has been published in premier venues across systems, security, and AI, including SIGMOD, VLDB, USENIX Security, MICRO, NeurIPS, IJCAI, ICCV, and EMNLP. Beyond his core focus, he actively collaborates across disciplines, contributing to projects in AI, machine learning, hardware systems, healthcare, and biomedicine. He received his Ph.D. in Computer Science from Duke University, where he was advised by Prof. Ashwin Machanavajjhala and Prof. Kartik Nayak.

With the recent rise of LLMs, it becomes increasingly possible (and necessary) to conduct human-subjects studies with AI tools. However, integrating AI in human-subjects research raises concerns about participant privacy, safety, and the confidentiality of research data. These concerns are exacerbated by the fact that many AI researchers have limited experience with human-subjects research, and most ethics review boards (e.g. IRBs) have little knowledge about AI. In this talk, I present findings from a series of investigations into human-centered AI research studies: our team systematically reviewed AI-related studies published at the ACM SigCHI conference, we interviewed researchers who conducted human-subjects studies with LLMs, and we conducted a scenario-based study to unpack study participants' opinions about AI-based research. We find that (1) many papers lack important details about basic study parameters, (2) researchers often fail to disclose the use of LLMs to research participants, and (3) participants are sensitive to study parameters like anonymization, data retention and consent, but are unaware of the threats of third-party data sharing and of data leakage through model training. I will discuss these findings, and more, during the talk. Speaker Bio: Dr. Bart Knijnenburg is an Associate Professor in Human-Centered Computing at Clemson University where he co-directs the Humans And Technology Lab (HATLab). His research explores the societal, social, and psychological aspects of human interaction with intelligent systems. His research has been funded by the NSF (including a CAREER award), the Department of Defense, the Department of Education, the Department of Justice, and corporate gifts.

Operational Technology (OT), when installed on an organization's network, becomes part of the overall cyber attack surface for an organization. When procuring this OT, it is important for the purchasing organization to understand how it will integrate with the existing network and security controls as well as understand what new risks it might introduce. The Trusted CI Operational Technology Procurement Vendor Matrix (the Matrix) provides a prioritized list of questions for organizations to send to manufacturers and suppliers to try to get as much of this information as possible. In this webinar, we will walk through what security issues impact OT, the role of procurement in mitigating security risks, our reasoning and process for developing the Matrix, and a walk through on how to use the Matrix at your organization. Questions and shared experiences with OT are encouraged. TARGET AUDIENCE: Organizational leadership, procurement department, IT, cybersecurity The Matrix can be found at https://trustedci.org/ot-matrix Speaker Bio: Chief Security Analyst Mark Krenz is focused on cybersecurity operations, research, and education. He has more than two decades of experience in system, network administration, programming, and system security and has spent the last decade focused on cybersecurity. He also serves as the CISO of Trusted CI.

In this talk, we will first discuss techniques to improve the resiliency of hazard monitoring systems. This includes optimizing machine learning training pipelines for wildfire detection to achieve faster, more accurate results while adapting to real-world constraints such as data variability and network latencies. We will also explore enabling multi-tenancy to maximize resource efficiency by allowing multiple hazard detection workflows to share infrastructure without compromising performance. Furthermore, we will present an in-depth analysis of power and energy consumption for edge devices deployed in remote and resource-constrained environments, emphasizing sustainable and scalable design choices that support long-term operation. Next, we will describe ongoing efforts to enhance the security of critical cyberinfrastructures. This includes developing techniques to prevent denial-of-service attacks that could disrupt hazard monitoring workflows and implementing secure data transmission mechanisms to safeguard information across distributed CI layers. Speaker Bios: Sudarsun Kannan is an Assistant Professor in the Computer Science Department at Rutgers University. His research focuses on operating system design and its intersection with computer architecture, distributed systems, and high-performance computing (HPC) systems. His work has been published in top venues such as ASPLOS, OSDI, and FAST, and he has received best paper awards at SOSP and ASPLOS, along with the Google Research Scholar award. He co-chaired the HotStorage'22 workshop and serves as an Associate Editor for ACM Transactions on Storage. Before joining Rutgers, he was a postdoctoral research associate at Wisconsin-Madison and graduated with an M.S. and Ph.D. from Georgia Tech. Ramakrishnan (Ram) Durairajan is an Associate Professor in the School of Computer and Data Sciences, and co-directs the Oregon Networking Research Group (ONRG) at the University of Oregon. Ram earned his Ph.D. and M.S. degrees in Computer Sciences from the University of Wisconsin - Madison and his B.Tech. in Information Technology from the College of Engineering, Guindy (CEG), Anna University. He has published over 50 peer-reviewed papers in various conferences, journals, and workshops. His research has been recognized with several awards including the NSF CAREER award, NSF CRII award, Ripple faculty fellowship, UO faculty research award, best paper awards from ACM CoNEXT and ACM SIGCOMM GAIA, and has been covered in several fora (NYTimes, MIT Technology Review, Popular Science, Boston Globe, Gizmodo, Mashable, among others). Recently, his research on Internet topology has been named as "One of the 100 Greatest Innovations," has been cited in FCC's Spectrum Frontiers 2d Report and Order, and has won a number of awards including the "Best of What's New" (in security category) by the Popular Science Magazine.

Data-driven collaborations often involve sharing large-scale datasets in cloud environments, where adversaries may exploit server vulnerabilities to access sensitive information. Traditional approaches, such as Trusted Execution Environments, lack the scalability for parallel processing, while techniques like homomorphic encryption incur prohibitive computational overheads. ARMOR addresses these limitations by developing encrypted querying techniques that support a variety of scientific data types and queries, balancing efficiency with privacy. The project’s interdisciplinary team focuses on advancing encryption methods, improving query performance for multidimensional data, and rigorously evaluating security risks and overheads under real-world scenarios. A recent research under ARMOR is the development of Secure Standard Aggregate Queries (SSAQ), a novel approach for secure aggregation on multidimensional sparse datasets stored on untrusted servers. Aggregation functions like SUM, AVG, COUNT, MIN, MAX, and STD are essential for scientific data analysis but pose privacy risks when performed on encrypted data. Existing methods using searchable encryption suffer from access pattern and volume leakage and are often limited to one-dimensional settings. SSAQ overcomes these challenges by employing d-dimensional segment trees to precompute responses for all possible query ranges, thus improving the efficiency of secure range queries. To further reduce leakage, SSAQ integrates Oblivious RAM (ORAM) to conceal data access patterns during query execution. This combination ensures a higher level of security, making SSAQ suitable for complex scientific data scenarios where sensitive information needs to be safeguarded. The approach significantly extends the applicability of searchable encryption techniques, offering a scalable and efficient solution for secure data analytics in cloud environments while minimizing privacy risks. Speaker Bio: Dr. Hoda Maleki is an Assistant Professor in the School of Computer and Cyber Sciences at Augusta University, specializing in system security, applied cryptography, and blockchain technology. She earned her Ph.D. in Computer Science and Engineering from the University of Connecticut. Dr. Maleki's research addresses critical security challenges, including IoT security, secure data retrieval in encrypted databases, and privacy-preserving data access in cloud environments. Her work leverages the Universally Composable (UC) security framework to analyze complex systems and employs multi-dimensional searchable encryption to protect massive scientific datasets. With over $1 million in NSF funding, her research advances scalable, efficient cryptographic solutions that meet the security needs of modern data-driven applications.

NSF cyberinfrastructure is undergoing a security transformation: a migration from X.509 user certificates to IETF-standard JSON Web Tokens (JWTs). This migration has facilitated a re-thinking of authentication and authorization among cyberinfrastructure providers: enabling federated authentication as a core capability, improving support for attribute, role, and capability-based authorization, and reducing reliance on prior identity-based authorization methods that created security and usability problems. In this webinar, members of the SciAuth project (https://sciauth.org/ - NSF award #2114989) will provide a short, hands-on tutorial for cyberinfrastructure professionals to learn about JWTs, including SciTokens (https://scitokens.org/ - NSF award #1738962). Participants will use Jupyter Notebooks to validate the security of JWTs and experiment with JWT-based authentication and authorization. Participants will gain an understanding of JWT basics suitable for understanding their security and troubleshooting any problems with their use. Speaker Bios: Dr. Jim Basney is a principal research scientist in the cybersecurity group at the National Center for Supercomputing Applications at the University of Illinois at Urbana-Champaign. He is the Director and PI of Trusted CI. Jim received his PhD in computer sciences from the University of Wisconsin-Madison. Dr. Derek Weitzel is a research assistant professor in the School of Computing at the University of Nebraska - Lincoln. He has been providing distributed computing solutions to the national cyberinfrastructures since 2009. He is a member of the OSG’s production operations team and leads the operations of the National Research Platform. His current areas of research involve distributed data management for shared and opportunistic storage, secure credential management, and network monitoring and analytics.

At ESnet, we pride ourselves on being cutting-edge, even if it causes a few scratches. Every new branch of Zeek is automatically built and tested in Gitlab CI. Then, every night, the latest successful 'master' build is deployed, along with all of our packages and scripts, to a test system via Ansible. As time permits, we roll out the latest build, in production, to over 40 servers. Through this process we've both been able to provide early feedback to the Zeek project about potential bugs and give ourselves an early warning system when changes impact our production plugins and scripts. Zeek is an open source network security monitoring tool. This does not focus on the use of Zeek itself, but rather the care and feeding of our installation footprint. Speaker Bio: Michael “Dop” Dopheide has spent the majority of his career working in the R&E community specializing in systems engineering, security research, incident response, and network intrusion detection. He especially enjoys helping coworkers debug problems at the packet and protocol levels. In addition to his operational security role, Dop helps support the open source Zeek community and volunteers every year to beta test the SANS Holiday Hack challenge.

Controls aren’t everything, but they are an important rubber-meets-the-road component of your cybersecurity strategy and program. This webinar will help you will understand the role controls play in a competent cybersecurity program through the lens of the Trusted CI Framework. And, with help cutting through the noise of the many, many controls and control sets in the wild, it will introduce you to the Transformative Twelve, a small, highly prioritized, evidence-based set of cybersecurity controls. Speaker Bio: Craig Jackson is Deputy Director at the Indiana University Center for Applied Cybersecurity Research, where his R&D interests include evidence-based approaches to security, cybersecurity fundamentals, and cybersecurity program development and governance. He leads collaborative work with critical infrastructure partners. His work includes the Trusted CI Framework, the Information Security Practice Principles, and the Cybertrack and USN’s PACT assessment methodologies. Craig’s education background is in law, education, psychology, and philosophy.

NSF's major facilities represent some of the most significant research facilities on the globe. The forthcoming revision to the Research Infrastructure Guide (or RIG) details NSF's guidance on securing these facilities and its expectations for cybersecurity programs at the major facilities. This presentation will explain how we approached shaping this guidance, the unique challenges we faced, and offer a peek at some of the resulting guidance the revised RIG will provide. Speaker Bio: Michael Corn has been a CISO at four institutions (UIUC, Illinois System, Brandeis University, and most recently UC San Diego). A regular author on a variety of privacy, cybersecurity and identity related topics, he is currently the Cybersecurity Advisor for Research Infrastructure in the Office of the Chief Officer for Research Facilities and additionally provides support to the Office of the Chief of Research Security Strategy and Policy within NSF. A recent online presentation on cybersecurity policy can be found at https://bit.ly/3JIpI8w.

Cybersecurity and privacy threats increasingly impact our daily lives, our national infrastructures, and our industry. Recent newsworthy attacks targeted nationally important infrastructure, our government, our researchers, and research facilities. The landscape of what needs to be protected and from what threats is rapidly evolving as new technologies are released and threat actors improve their capabilities through experience and close collaboration. Meanwhile, defenders often work in isolation, use private data and facilities, and produce defenses that are quickly outpaced by new threats. To transform cybersecurity and privacy research into a highly integrated, community-wide effort, researchers need a common, rich, representative research infrastructure that meets the needs across all members of the community, and facilitates reproducible science. To meet these needs, USC Information Sciences Institute and Northeastern University have been funded by the NSF mid-scale research infrastructure program to build Security and Privacy Heterogeneous Environment for Reproducible Experimentation (SPHERE). This infrastructure will offer access to an unprecedented variety of hardware, software, and other resources connected by user-configurable network substrate, and protected by a set of security policies uniquely aligned with cybersecurity and privacy research needs. SPHERE will offer six user portals, closely aligned with needs of different user groups. It will support reproducible research through a combination of infrastructure services (easy experiment packaging, sharing and reuse) and community engagement activities (development of realistic experimentation environments and contribution of high-quality research artifacts). Speaker Bios: Dr. Jelena Mirkovic is Principal Scientist at USC-ISI and Research Associate Professor at USC. She received her MS and PhD from UCLA, and her BSc from University of Belgrade, Serbia. Jelena's research interests span networking and cybersecurity fields, as well as testbed experimentation. Her current research is focused on authentication, use of machine learning for network attack detection, large-scale dataset labeling for security, and user privacy. She is the lead PI on the SPHERE project. Mr. David Balenson is Senior Supervising Computer Scientist and Associate Director of the Networking and Cybersecurity Division at USC-ISI. He received his MS and BS in Computer Science from the University of Maryland. His current research interests include cybersecurity and privacy for critical infrastructure and cyber-physical systems including automotive and autonomous vehicles, experimentation and test, technology transition, and multidisciplinary research. He is the Community Outreach Director for SPHERE.

The ACCORD cyberinfrastructure project at the University of Virginia (UVA) successfully developed and deployed a community infrastructure providing access to secure research computing resources for users at underserved, minority-serving, and non-PhD-granting institutions. ACCORD's operational model is built around balancing data protection with accessibility. In addition to providing secure research computing resources and services, key outcomes of ACCORD include creation of a set of policies that enable researchers external to UVA to access and use ACCORD. While the ACCORD expedition achieved its technical and operational goals, its broader mission of broadening access to underserved users had limited success. Toward gaining a better understanding of the barriers to researchers accessing ACCORD, our team carried out two community outreach efforts to engage with researchers and computing service leaders to hear their pain points as well as solicit their input for an accessible community infrastructure. In this talk, we will describe the ACCORD infrastructure and its operational model. We will also discuss insights from our effort to develop policies to balance accessibility with security. And finally, we wil share lessons learned from community outreach efforts to understand institutional and social barriers to access. Speaker Bios: Ron Hutchins: In the early 1980’s, Ron worked at the Georgia Institute of Technology to create a networking laboratory in the College of Computing teaching data communications courses there. After moving to the role of Director of Campus Networks in 1991, Ron founded and led the Southern Crossroads network aggregation (SoX) across the Southeast. In 2001 after receiving his PhD in computer networks, he took on the role of Chief Technology Officer for the campus. In August of 2015, Ron moved into the role of Vice President of Information Technology for the University of Virginia, working to build partnerships across the campus. Recently, Ron has moved from VP to research faculty in the Computer Science department at UVA and is participating broadly across networking and research computing in general including work with the State of California building out the broadband fiber network backbone across the state. Tho Nguyen is a computer science and policy expert. He served as project manager for the ACCORD effort from 2019-2021, and continues to support the project implementation and growth. Nguyen is currently a Senior Program Officer at the National Academies of Sciences, Engineering, and Medicine. From 2015-2021 Nguyen was on the research staff in the Department of Computer Science at the University of Virginia where he worked on compute-in-memory and developing HPCs for research. Prior to UVA, he was a AAAS Science and Technology Policy Fellow at the National Science Foundation where he worked primarily on the Cyber Physical Systems program. Nguyen holds a PhD in Systems & Controls (Electrical Engineering) from the University of Washington.

The envisioned advantage of sharing research data lies in its potential for reuse. Although many scientific disciplines are embracing data sharing, some face constraints on the data they can share and with whom. It becomes crucial to establish a secure method that efficiently facilitates sharing and verification of data and metadata while upholding privacy restrictions to enable the reuse of scientific data. This presentation highlights our NSF-funded Open Science Chain (OSC) project, accessible at https://www.opensciencechain.org. Developed using blockchain technologies, the OSC project aims to address challenges related to the integrity and provenance of research artifacts. The project establishes an API-based data integrity verification management service for data-driven research platforms and hubs, aiming to minimize data information loss and provide support for managing diverse metadata standards and access controls. Speaker Bio: Subhashini Sivagnanam is the manager of the Cyberinfrastructure Services and Solutions (CISS) group at the San Diego Supercomputer Center/ UCSD. Her research interests predominantly lie in distributed computing, cyberinfrastructure development, scientific data management, and reproducible science. She serves as the PI/Co-PI on various NSF/NIH projects related to scientific data integrity and developing cyberinfrastructure software. Furthermore, she oversees the management of UC San Diego’s campus research cluster known as the Triton Shared Computing Cluster.

As the use of wastewater for public health surveillance continues to expand, inevitably sample collection will move from centralized wastewater treatment plants to sample collection points within the sewer collection system to isolate individual neighborhoods and communities. Collecting data at this geospatial resolution will help identify variation in select biomarkers within neighborhoods, ultimately making the wastewater-derived data more actionable. However a challenge in achieving this is the nature of the wastewater collection system, which aggregates and commingles wastewater from various municipalities. Thus various stakeholders from different cities must collectively provide information to separate wastewater catchments to achieve neighborhood-specific public health information. Data sharing restrictions and the need for anonymity complicates this process. This talk presents our approaches to enabling data privacy in wastewater-based epidemiology. Our methodology is built upon a cryptographic technique, Homomorphic Encryption (HE), ensuring privacy. Additionally, we outline a technique to enhance the performance of HE, which could be of independent interest. Speaker Bio: Ni Trieu is currently an Assistant Professor at Arizona State University (ASU). Her research interests lie in the area of cryptography and security, with a specific focus on secure computation and its applications such as private set intersection, private database queries, and privacy-preserving machine learning. Prior to joining ASU, she was a postdoc at UC Berkeley. She received her Ph.D. degree from Oregon State University.