UMBC CSEE Research Day 2024 Program

Computer Science and Electrical Engineering
University of Maryland, Baltimore County

9 am to 4 pm, Friday, May 3, 2024
ITE Building, 3rd Floor
Talks online via WebEx

A celebration of selected research accomplishments from 2023–2024 by UMBC faculty and students in the Department of Computer Science and Electrical Engineering.  Open to faculty, researchers, students, and visitors.

The reception, opening session, lunch, poster sessions, and awards will take place in the 3rd floor hallways of the ITE Building.  Talks will be broadcast online via WebEx from ITE 325B.

WebEX Link: Talks online via WebEx

 

Coffee and Tea available all day in the Kitchen Area of ITE 325

Opening Remarks (9:20am–9:30am), Mohamed Younis, Chair, Dept. of CSEE

Session I (9:30am–10:40am) – Each talk is 20 mins

9:30am – Ram P Rustagi,

      • Title: Experiential Learning of Computer Networks
      • Abstract: The CSEE/DoIT Cyber Range aims to provide a sandbox environment where students can conduct experiments and carry out hands-on exercises primarily related to Computer Networks and Security as well as in other areas of Computer Science. Exercises related to securities are conducted on several Virtual Machines (VMs) comprising of an attacker system, victim system, and other VMs as needed. The exercises related to Computer Networks are mostly carried out using docker instances. Use of docker instances consumes much less compute resources and thus can scale up to support a large number of users. Currently, Cyber Range is being used in CMSC-481, CMSC-421, and independent studies and is expected to be used by students in other courses. This talk provides a brief overview of Computer Networks Experiential Learning exercises carried out by students CMSC-481 (both in Fall 2023 and Spring 2024).

9:55am – Austin Murdock (invited speaker),

        • Title: Automated Cyber Defense (Invited Speaker)
        • Abstract: In this talk, Dr. Murdock will discuss a few of the main challenges in cybersecurity today and introduce Computational Mapping and its applications to Automated Cyber Defense.
        • Bio: Dr. Austin Murdock is the inventor of Computational Mapping and a subject matter expert in Cybersecurity, Networking, High Performance Computing, and Machine Learning. He invented Computational Mapping based on his career as a cybersecurity researcher at the International Computer Science Institute, University of California, Berkeley, Massachusetts Institute of Technology, Johns Hopkins University, and the Department of Defense.

10:20am – Sai Vallurupalli,

      • Title: Examining Participant-specific Goals for a Deeper Understanding of Complex Events (PhD student award)
      • Abstract:Interpreting and assessing goal driven actions is vital to understanding and reasoning over complex events. It is important to be able to acquire the knowledge needed for this understanding, though doing so is challenging. We argue that such knowledge can be elicited through a participant achievement lens. We analyze a complex event in a narrative according to the intended achievements of the participants in that narrative, the likely future actions of the participants, and the likelihood of goal success. We collect 6.3K high quality goal and action annotations reflecting our proposed participant achievement lens, with an average weighted Fleiss-Kappa IAA of 80%;. Our collection contains annotated alternate versions of each narrative. These alternate versions vary minimally from the “original” story, but can license drastically different inferences. Our findings suggest that while modern large language models can reflect some of the goal-based knowledge we study, they find it challenging to fully capture the design and intent behind concerted actions. We show that smaller models fine-tuned on our dataset can achieve performance surpassing larger models.

Break (10:40am–11:00am)

Student Poster Session (11:00am–12:00pm) – faculty vote for the three best posters using Range Voting

Free Lunch (12:00pm–1:00pm) – pick up from the Kitchen Area of ITE 325 and enjoy in the open spaces around the 3rd floor of ITE as well as the 2nd floor patio, weather permitting

Session II (1:00pm–2:10pm)

1:00pm – Ajinkya Borle,

      • Title: Biclustering a dataset using photonic quantum computing
      • Abstract: Biclustering is a problem in machine learning and data mining that seeks to group together rows and columns of a dataset according to certain criteria. In this work, we highlight the natural relation that quantum computing models like boson and Gaussian boson sampling (GBS) have to this problem. We first explore the use of boson sampling to identify biclusters based on matrix permanents. We then propose a heuristic that finds clusters in a dataset using Gaussian boson sampling by (i) converting the dataset into a bipartite graph and then (ii) running GBS to find k-densest subgraphs. Our simulations for the above proposed heuristics show promising results for future exploration in this area.

1:25pm – Charles Nicholas,

      • Title: Malicious Needles in Benign Haystacks
      • Abstract: We are addressing the problem of identifying malicious code (malcode) that may be hidden in large, but otherwise benign software systems.  This is not a new problem, and current anti-virus systems can search for specific patterns of binary strings that indicate the presence of malcode.  Our approach involves the use of a compression-based fuzzy similarity metric called LZJD.  Over the last year or two, we have explored the problem of searching large executable binaries for instances of malcode.  Although it may make sense to break the binaries up into component functions, and search within those, the process of finding those functions takes time, and may not be effective.  In this work, we restrict ourselves to working with executable binaries.  We assume that we have a number of malware specimens, in PE Format, and a large “suspect” binary, which we wish to search for malcode. For comparing segments of executable binaries, we know of two viable approaches. One approach involves the use of n-grams, and the other involves a compression-based similarity metric such as LZJD In this talk we compare these n-gram and compression-based approaches on a set of malware specimens, and a large but benign executable binary.
        In memory of our friend and colleague Dhruvil Modi, 2001-2024.

1:50pm – Akash Vartak,

      • Title: DeBUGCN – Detecting Backdoors in CNNs Using Graph Convolutional Networks (PhD student award)
      • Abstract: Deep neural networks (DNNs) are becoming commonplace in critical applications, making their susceptibility to backdoor (trojan) attacks a significant problem. In this paper, we introduce a novel backdoor attack detection pipeline, detecting attacked models using graph convolution networks (DeBUGCN). We use the static weights of a DNN’s layers to convert the layer structures to graphs. The GCN is then used as a binary classifier on these graphs, yielding a trojan or clean determination for the DNN. When we compare our results on several datasets with state-of-the-art trojan detection algorithms, DeBUGCN exhibits comparable accuracy with less computation time.

Break (2:10pm–2:30pm)

Session III (2:30pm–3:40pm)

2:30pm – Lara Martin,

      • Title: Bridging the Social & Technical Divide in Augmentative and Alternative Communication (AAC) Applications for Autistic Adults
      • Abstract: Natural Language Processing (NLP) techniques are being used more frequently to improve high-tech Augmentative and Alternative Communication (AAC), but many of these techniques are integrated without the inclusion of the users’ perspectives. As many of these tools are created with children in mind, autistic adults are often neglected in the design of AAC tools to begin with. We conducted in-depth interviews with 12 autistic adults to find the pain points of current AAC and determine what general technological advances they would find helpful. We found that in addition to technological issues, there are many societal issues as well. We found 9 different categories of themes from our interviews: input options, output options, selecting or adapting AAC for a good fit, when to start or swap AAC, benefits (of use), access (to AAC), stumbling blocks for continued use, social concerns, and lack of control. In this paper, we go through these nine categories in depth and then suggest possible guidelines for the NLP community, AAC application makers, and policy makers to improve AAC use for autistic adults.

2:55pm – Roberto Yus,

      • Title: Co-zyBench: Using Co-Simulation and Digital Twins to Benchmark Thermal Comfort Provision in Smart Buildings.
      • Abstract: Heating, Ventilation, and Air Conditioning (HVAC) systems account for 40% to 50% of energy usage in commercial buildings. Thus, innovative ways to control and manage HVAC systems while preserving occupants’ comfort are required. State-of-the-art solutions employ pervasive systems with sensors or smart devices to gauge individual thermal sensations, yet assessing these methods is challenging. Real-world experiments are expensive, limited in access, and often overlook occupant and regional diversity. To address this, we introduce Co-zyBench, a benchmark tool using a Digital Twin (DT) approach for evaluating personalized thermal comfort systems. It employs a co-simulation middleware interfacing between a DT of the smart building and its HVAC system and another DT representing occupants’ dynamic thermal preferences in various spaces. The DTs that support Co-zyBench are generated based on information, including data captured by sensors, of the space in which the thermal comfort system has to be evaluated. Co-zyBench incorporates metrics for energy consumption, thermal comfort, and occupant equality. It also features reference DTs based on standard buildings, HVAC systems, and occupants with diverse thermal preferences.

3:20pm – Aamir Hamid,

      • Title: GenAIPABench A Benchmark for Generative AI-based Privacy Assistants (PhD student award)
      • Abstract: Privacy policies of websites are often lengthy and intricate. Privacy assistants assist in simplifying policies and making them more accessible and user-friendly. The emergence of generative AI (genAI) offers new opportunities to build privacy assistants that can answer users’ questions about privacy policies. However, genAI’s reliability is a concern due to its potential for producing inaccurate information. This study introduces GenAIPABench, a benchmark for evaluating Generative AI-based Privacy Assistants (GenAIPAs). GenAIPABench includes: 1) A set of questions about privacy policies and data protection regulations, with annotated answers for various organizations and regulations; 2) Metrics to assess the accuracy, relevance, and consistency of responses; and 3) A tool for generating prompts to introduce privacy documents and varied privacy questions to test system robustness. We evaluated three leading genAI systems—ChatGPT-4, Bard, and Bing AI—using GenAIPABench to gauge their effectiveness as GenAIPAs. Our results demonstrate significant promise in genAI capabilities in the privacy domain while also highlighting challenges in managing complex queries, ensuring consistency, and verifying source accuracy.

      Student Poster Awards (3:40-4:00pm)

      Adjourn (4:00pm)