talk: Forward & Inverse Causal Inference in a Tensor Framework, 1-2 pm ET, 3/29

Forward and Inverse Causal Inference in a Tensor Framework

M. Alex O. Vasilescu
Institute of Pure and Applied Mathematics, UCLA

1-2:00 pm Monday, March 29, 2021
via WebEx

Developing causal explanations for correct results or for failures from mathematical equations and data is important in developing a trustworthy artificial intelligence, and retaining public trust.  Causal explanations are germane to the “right to an explanation” statute, i.e., to data-driven decisions, such as those that rely on images.  Computer graphics and computer vision problems, also known as forward and inverse imaging problems, have been cast as causal inference questions consistent with Donald Rubin’s quantitative definition of causality, where “A causes B” means “the effect of A is B”, a measurable and experimentally repeatable quantity. Computer graphics may be viewed as addressing analogous questions to forward causal inference that addresses the “what if” question, and estimates a change in effects given a delta change in a causal factor. Computer vision may be viewed as addressing analogous questions to inverse causal inference that addresses the “why” question which we define as the estimation of causes given a forward causal model, and a set of observations that constrain the solution set.  Tensor algebra is a suitable and transparent framework for modeling the mechanism that generates observed data.  Tensor-based data analysis, also known in the literature as structural equation modeling with multimode latent variables, has been employed in representing the causal factor structure of data formation in econometrics, psychometric, and chemometrics since the 1960s.  More recently, tensor factor analysis has been successfully employed to represent cause-and-effect in computer vision, and computer graphics, or for prediction and dimensionality reduction in machine learning tasks.   

M. Alex O. Vasilescu received her education at the Massachusetts Institute of Technology and the University of Toronto. She is currently a senior fellow at UCLA’s Institute of Pure and Applied mathematics (IPAM) that has held research scientist positions at the MIT Media Lab from 2005-07 and at New York University’s Courant Institute of Mathematical Sciences from 2001-05.  Vasilescu introduced the tensor paradigm for computer vision, computer graphics, and machine learning. She addressed causal inferencing questions by framing computer graphics and computer vision as multilinear problems. Causal inferencing in a tensor framework facilitates the analysis, recognition, synthesis, and interpretability of data. The development of the tensor framework has been spearheaded with premier papers, such as Human Motion Signatures (2001), TensorFaces (2002), Multilinear Independent Component Analysis (2005), TensorTextures (2004), and Multilinear Projection for Recognition (2007, 2011). Vasilescu’s face recognition research, known as TensorFaces, has been funded by the TSWG, the Department of Defenses Combating Terrorism Support Program, Intelligence Advanced Research Projects Activity (IARPA), and NSF. Her work was featured on the cover of Computer World and in articles in the New York Times, Washington Times, etc. MIT’s Technology Review Magazine named her to their TR100 list of honorees, and the National Academy of Science co-awarded the Keck Futures Initiative Grant.  

ACM career talk: career opportunities in data privacy

Continuing with our Innovation, Collaboration, Job Search, and Career help theme, the ACM UMBC chapter is back again, hosting another session on the coming Friday with Sameer Ahirrao, a Founder and CEO of Ardent Privacy. He will be talking about Innovation, Collaboration, and Career Opportunities in Data Privacy. Find more on how you can get a part-time off-campus or full-time internship under MIPS (Maryland Industrial Partnership) Program with Ardent Privacy.

Join us for insights from him and a Q&A session with Sameer.
See you on Friday, March 16 at 3:00 pm EST on WebeX.  For more information, contact:  .

talk: Transparent Dishonesty: Front-Running Attacks on Blockchain, 12-1 pm ET 3/26

The UMBC Cyber Defense Lab presents

Transparent Dishonesty: Front-Running Attacks on Blockchain

Professor Jeremy Clark
Concordia Institute for Information Systems Engineering
Concordia University, Montreal, Canada

12–1 pm ET Friday, March 26, 2021
online via WebEx

We consider front-running to be a course of action where an entity benefits from prior access to privileged market information about upcoming transactions and trades. Front-running has been an issue in financial instrument markets since the 1970s. With the advent of blockchain technology, front-running has resurfaced in new forms we explore here, instigated by blockchain’s decentralized and transparent nature. I will discuss our “systemization of knowledge” paper which draws from a scattered body of knowledge and instances of front-running across the top 25 most active decentral applications (DApps) deployed on Ethereum blockchain. Additionally, we carry out a detailed analysis of initial coin offering (ICO) and show evidence of abnormal miner’s behavior indicative of front-running token purchases. Finally, we map the proposed solutions to front-running into useful categories.

Jeremy Clark is an associate professor at the Concordia Institute for Information Systems Engineering. At Concordia, he holds the NSERC/Raymond Chabot Grant Thornton/Catallaxy Industrial Research Chair in Blockchain Technologies. He earned his Ph.D. from the University of Waterloo, where his gold medal dissertation was on designing and deploying secure voting systems including Scantegrity—the first cryptographically verifiable system used in a public sector election. He wrote one of the earliest academic papers on Bitcoin, completed several research projects in the area, and contributed to the first textbook. Beyond research, he has worked with several municipalities on voting technology and testified to both the Canadian Senate and House finance committees on Bitcoin. email:

Host: Alan T. Sherman, . Support for this event was provided in part by the National Science Foundation under SFS grant DGE-1753681. The UMBC Cyber Defense Lab meets biweekly Fridays. All meetings are open to the public. Upcoming CDL Meetings: April 9, (UMBC), MeetingMayhem: A network adversarial thinking game; April 23, Peter Peterson (University of Minnesota Duluth), Adversarial thinking;
May 7, Farid Javani (UMBC), Anonymization by oblivious transfer.

talk: Machine Learning: New Methodology for Physical & Social Sciences, 1pm ET 3/24

24 hour LIDAR backscatter profiles and PBLH points generated from image machine learning system

The Infusion of Machine Learning as a New Methodology for the Physical and Social Sciences

Dr. Jennifer Sleeman

1:00-2:00 pm ET, Wednesday, March 24
Online via WebEx

Machine learning has made improvements in many areas of computing. Recently attention has been given to infusing social science methodology with machine learning. In addition, the physical sciences have begun to embrace machine learning to augment their physical parameterization and to discover new features in their computations. I will describe my work that relates to these new emerging areas of research. I will first describe our machine learning research efforts related to understanding the changing role of climate and its effects on society. I will describe how this methodology was also applied to understanding cyber-related exploits. As part of this work, I developed an expertise in generative modeling, which led to a patent in generative and translation-based methods applied to imagery. These ideas were fundamental to a contribution in machine learning using quantum annealing. Quantum computing holds promise for deep learning to reach model convergence faster than classical computers. I will describe work related to developing a new hybrid method that overcame qubit limitations for image generation. 

In addition, I will describe my current work related to machine learning for the Physical Sciences. As part of a multi-disciplinary team from UMBC and other universities, my current work explores ways to augment and replace existing physical parameterizations with neural network based models. I have led a research effort to calculate the planetary boundary layer’s height (PBLH) used for ceilometer-based backscatter profiles and satellite-borne lidar instruments. This work addresses the largest uncertainty in climate change, namely the role of aerosols (dust, carbon, sulfates, sea salt, etc.). We employ a novel method that includes a deep segmentation neural network that uses near-time continuous profiles forming an image to determine boundary layer heights. This method overcomes limitations in wavelet approaches which are unable to identify the PBLH under certain conditions. I will also give a preview of two efforts related to Long Short Term Memory (LSTM) neural networks related to learning PBLH changes over time. These research efforts result from collaborations with two students in the UMBC CSEE department and are being published and presented at the AAAI 2021 Spring Symposium on Combining Artificial Intelligence and Machine Learning with Physics Sciences. 

Dr. Jennifer Sleeman is a Research Assistant Professor in Computer Science at the University of Maryland, Baltimore County (UMBC). Her research interests include generative models, natural language processing, semantic representation, image generation, and deep learning. Dr. Sleeman received the prestigious recognition of being a 2019 EECS Rising Star. She was also recognized in 2017 as one of the best Data Scientists in the Washington, DC region by DCFemTech. She defended her Ph.D. thesis, Dynamic Data Assimilation for Topic Modeling (DDATM) in 2017 under Tim Finin and Milton Halem. Her thesis-related work was awarded a Microsoft “AI for Earth” resource grant in 2017 and 2018 and also won the best paper award in the Semantic Web for Social Good Workshop presented at International Semantic Web Conference in 2018. She was an invited guest panelist at the AI for Social Good AAAI Fall Symposium in 2019 and was also an invited keynote speaker at the Sixth IEEE International Conference on Data Science and Engineering (ICDSE 2020), where she presented her ideas related to AI for Social Good and Science. She is an active research scientist in generative deep learning methods for which she holds a patent. She has over 12 years of machine learning experience and over 22 years of software engineering experience, in both academic and government/industry settings. She is currently funded by NASA and NOAA (PI). She also teaches Introduction to Artificial Intelligence at the University of Maryland, Baltimore County (UMBC) and currently mentors two Master’s students

talk: (Don’t) Mind the Gap: Bridging the Worlds of People and IoT Devices, 1-2 ET 3/23

TIPPERS is an IoT data management middleware system developed at UCI that manages IoT smart spaces by collecting sensor data, inferring semantically meaningful information, and providing developers with data for intelligent applications.

(Don’t) Mind the Gap: Bridging the Worlds of People and IoT Devices

Dr. Roberto Yus
University of California, Irvine

1:00-2:00 pm ET, Tuesday, 23 March 2021
online via WebEx

The Internet of Things (IoT) has the potential to improve our lives through different services given the diversity of smart devices and their capabilities. For example, the IoT can empower services to make the re-opening of business during the current pandemic safer by monitoring adherence to regulations. But the large amounts of highly heterogeneous data captured by IoT devices typically require further processing to become useful information. The challenge is thus for IoT systems to determine which sensor data has to be captured/stored/processed/shared to, for instance, determine the occupancy of a specific office building or the spaces in which a potential exposure took place. This becomes even more challenging when IoT systems have to take into account the privacy preferences of individuals, such as the need to prevent sharing data about their daily patterns or habits.

In this talk, I will discuss my efforts into helping IoT systems bridge the gap between the world of IoT devices and the world where people act. First, I will introduce a model to represent knowledge about sensors/actuators, people, spaces, events, and their relationships. Based on the model, I will explain an algorithmic solution to translate user requests and privacy preferences defined in a high-level, more semantically meaningful way into operations on IoT devices and their captured data. Second, I will talk about the enforcement of privacy preferences in the context of the IoT. Finally, I will overview my experience building and deploying an IoT data management system, TIPPERS, which has been deployed at UC Irvine and two US Navy vessels and is soon to be deployed on other campuses. I will conclude the talk by discussing the exciting future work opportunities towards supporting the next generation of ubiquitous IoT data management systems and technologies that autonomously, transparently, and at scale, balance the trade-off between providing users with high utility and respecting people’s privacy requirements.

Roberto Yus is a Postdoctoral Researcher in the Computer Science department at the University of California, Irvine working with Prof. Sharad Mehrotra. Before that, he spent a year as a visiting researcher at the University of Maryland, Baltimore County working with Prof. Anupam Joshi and Prof. Tim Finin. He obtained his Ph.D. in Computer Science from the University of Zaragoza, Spain, funded through a 4-year fellowship from the Spanish Ministry of Science and Innovation. His research interests are in the fields of data management, knowledge representation, privacy, and the Internet of Things (IoT). His research focuses on the design of semantic data management solutions to empower IoT systems to understand user information requirements and user privacy preferences and adapt their operations taking those into account. Roberto’s research has been published in top-tier conferences and journals such as VLDB and the Journal of Web Semantics. He is part of the editorial board of the “Sensors” and “Frontiers in Big Data” journals and has served as part of the organizing and program committee of several conferences and workshops in addition to serving as an external reviewer for multiple conferences and journals.

talk: Towards Contextual Security of AI-enabled platforms, 1-2 pm ET 3/22

Towards Contextual Security of AI-enabled platforms

Dr. Nidhi Rastogi
Rensselaer Polytechnic Institute

1-2:00pm ET, Monday, 22 March 2021

via WebEx

The explosive growth of Internet-connected and AI-enabled devices and data produced by them has introduced significant threats. For example, malware intrusions (SolarWinds) have become perilous and extremely hard to discover, while data breaches continue to compromise user privacy (Zoom credentials exposed) and endanger personally identifiable information. My research takes a holistic approach towards systems and platforms to address security-related concerns using contextual and explainable models. 

In this talk, I will present ongoing work that analyzes and improves the cybersecurity posture of Internet-connected systems and devices using automated, trustworthy, and contextual AI-models. Specifically, my research in malware threat intelligence gathers diverse information from varied datasets – system and network logs, source code, and text. In [1], an open-source ontology (MALOnt) contextualizes threat intelligence by aggregating malware-related information into classes and relations. TINKER [2, 3] – the first open-source malware knowledge graph, instantiates MALOnt classes and enables information extraction, reasoning, analysis, detection, classification, and cyber threat attribution. At present, the research is addressing the trustworthiness of information sources and extractors.

1. RastogiN., Dutta, S., Zaki, M. J., Gittens, A., & Aggarwal, C. (2020). MALOnt: An ontology for malware threat intelligence, In KDD’20 Workshop at International workshop on deployable machine learning for security defense. Springer, Cham.

2. RastogiN., Dutta, S., Christian, R., Gridley, J., Zaki, M. J., Gittens, A., and Aggarwal, C.  (2021). Knowledge graph generation and completion for contextual malware threat intelligence. In USENIX Security’21, Accepted.

3. Yee, D., Dutta, S., RastogiN., Gu, C., and Ma, Q. (2021). TINKER: Knowledge graph for threat intelligence. In ACL- IJCNLP’21, Under Review.

Dr. Nidhi Rastogi is a Research Scientist at Rensselaer Polytechnic Institute. Her research is at the intersection of cybersecurity, artificial intelligence, large-scale networks, graph analytics, and data privacy. She has papers accepted at top venues such as USENIX, TrustCom, ISWC, Wireless Telecommunication Symposium, and Journal of Information Policy. For the past two years, Dr. Rastogi has been the lead PI for three cybersecurity, privacy research projects and a contributor to one healthcare AI project. For her contributions to cybersecurity and encouraging women in STEM, Dr. Rastogi was recognized in 2020 as an International Women in Cybersecurity by the Cyber Risk Research Institute. She was a speaker at the SANS cybersecurity summit and the Grace Hopper Conference. Dr. Rastogi is the co-chair for DYNAMICS workshop (2020-) and has served as a committee member for DYNAMICS’19, IEEE S&P’16 (student PC), invited reviewer for IEEE Transactions on Information Forensics and Cybersecurity (2018,19), FADEx laureate for the 1st French-American Program on Cyber-Physical Systems’16, Board Member (N2Women 2018-20), and Feature Editor for ACM XRDS Magazine (2015-17). Before her Ph.D. from RPI, Dr. Rastogi also worked in the industry on heterogeneous wireless networks (cellular, 802.1x, 802.11) and network security through engineering and research positions at Verizon and GE Global Research Center, and GE Power. She has interned at IBM Zurich, BBN Raytheon, GE GRC, and Yahoo, which provides her a quintessential perspective in applied industrial research and engineering.

UMBC expands live online peer tutoring to include computing courses

Amanda Knapp speaks with a student in the Academic Success Center. Photo by Marlayna Demond ’11 for UMBC.

UMBC expands live online peer tutoring to include computing courses

When Amanda Knapp heard last fall from Anupam Joshi, professor and chair of computer science and electrical engineering (CSEE), that his department wanted to offer online tutoring to students in their courses, she was ready to help make it happen. COVID or no COVID, she says, “It just made sense.”

Knapp is associate vice provost and assistant dean of Undergraduate Academic Affairs, and she manages UMBC’s Academic Success Center (ASC). “We already had an established system in place and could provide the administration, training, staffing, and assessment,” she explains. “CSEE provided the funding and identified potential tutors for a variety of computer science courses, and we did the rest.”

Just a few months after the partnership began, it expanded to include courses in the information systems (IS) department. And this semester, tutors are supporting ten IS computing courses.

In total, the ASC’s new Computing Success Center now offers tutoring for 21 courses from across the College of Engineering and Information Technology (COEIT). And the center continues to grow, supported by COEIT Dean Keith J Bowman and Dean Katharine Cole, Undergraduate Academic Affairs.

Peer tutors help students of any major, from computing fields to the arts and life sciences, learn coding and other computing skills.

“The ethos of UMBC is to share knowledge and collaborate with others instead of being proprietary,” says Helena Mentis, associate professor of information systems. Mentis is COEIT’s associate dean of academic programs and learning, and one of the College leads on the partnership.

“At the end of the day,” she says, “our goal at UMBC is to ensure that there are multiple pathways for students to get the assistance they need to succeed and remove any barriers for access to help.”

You can read more about UMBC’s new Computing Success Center in this UMBC News article.

Adapted from a UMBC News article by Gregory J. Alexander and Dinah Winnick

UMBC CSEE Newsletter, Spring 2021

csee spring newsletter

UMBC CSEE Newsletter, Spring 2021

The UMBC CSEE department’s Spring 2021 newsletter is available with recent news on our faculty, students, alumni, and programs.

ACM chapter talk: Career, job search, and interviewing tips, 4-5 pm Sat, 3/13

ACM chapter talk
Career, job search, and interviewing tips

Nikhil Kumar Mengani (UMBC MS CS ’18), Microsoft SDE

The UMBC student ACM chapter will hold a session on careers and job searches from 4:00 pm to 5:00 pm ET on Saturday, March 13.  Nikhil Mengani, a UMBC graduate and current Microsoft Software Development Engineer, will talk about interview tips, using LinkedIn, and overall job search best practices. 

Join the online meeting for some great insights and a Q&A session with Nikhil.  Join via webex. For more information, contact Samit Shivadekar at .

talk: EIPC: Efficient Asynchronous BFT with Adaptive Security, 12-1 Fri 3/12

The UMBC Cyber Defense Lab presents

EIPC: Efficient Asynchronous BFT with Adaptive Security

Chao Liu, CSEE, UMBC

12:00–1:00 pm ET, Friday, 12 March 2021
via WebEx

We present EPIC, a novel and efficient asynchronous Byzantine fault-tolerant (BFT) protocol with adaptive security. We characterize efficient BFT protocols using adaptive vs. static corruptions corruption models. EPIC takes a new approach to adaptively secure asynchronous BFT. It uses the adaptively secure threshold pseudorandom function (PRF) scheme for coin tossing and uses the Cobalt asynchronous binary agreement (ABA) protocol, which resolves the liveness issue of HoneyBadgerBFT and BEAT. As our new protocol modifies almost all building blocks for asynchronous BFT (including ABA, threshold PRF, and threshold encryption but not Byzantine reliable broadcast (RBC)), evaluating which component dominates the performance bottleneck is a difficult task. We mix and match different building blocks to implement four asynchronous BFT protocols and evaluate their performance. Via a five-continent deployment on Amazon EC2, we show that EPIC is slightly slower for small and medium-sized networks than the most efficient asynchronous BFT protocols with static security. We also find when the number of replicas less than 46, EPIC’s throughput is stable, achieving a peak throughput of 8,000–12,500 tx/sec using t2.medium VMs. When the network size grows larger, EPIC is not as efficient as those with static security, with a throughput of 4,000–6,300 tx/sec.

BFT state machine replication is the only known software solution for masking arbitrary failures and malicious attacks. BFT has been regarded as the model for building permissioned blockchains, where the distributed ledgers (i.e., replicas) know each other’s identities but may not trust each other.

Asynchronous protocols are inherently more robust against timing and denial-of-service (DoS) attacks. Two recent asynchronous BFT systems—HoneyBadgerBFT proposed by Miller et al. in CCS’16 and BEAT by Duan et al. in CCS’18—have comparable performance as partially synchronous BFT protocols and can scale to 100 replicas. The protocols, however, achieve static security, where the adversary needs to choose the set of corrupted replicas before protocol execution. This security property is weaker than that for many existing BFT protocols (e.g., PBFT), which achieve adaptive security, where the adversary can choose to corrupt replicas at any moment during the execution of the protocol.

Chao Liu is a Ph.D. candidate in computer science at UMBC, working with Alan Sherman. His research interests focus on cryptography, cybersecurity, and distributed systems.

Host: Alan T. Sherman, . Support for this event was provided in part by the National Science Foundation under SFS grant DGE-1753681. The UMBC Cyber Defense Lab meets biweekly Fridays. All meetings are open to the public. Upcoming CDL Meetings include Mar 26, Jeremy Clark (Concordia); April 9, (UMBC), MeetingMayhem: A network adversarial thinking game; April 23, Peter Peterson (University of Minnesota Duluth), Adversarial thinking; and May 7, Farid Javani (UMBC), Anonymization by oblivious transfer

1 2 3 4 5 142