Sabrina Perry
Assistant Professor
Computer Science and Information Technology
Austin Peay State University
Biography
Dr. Sabrina Perry is an Assistant Professor of Computer Science and Information Technology at Austin Peay State University. She teaches courses ranging from introductory programming in Java to advanced topics in cybersecurity, including Security Strategies in Windows and Linux Platforms and Securing Cyberspace. Her teaching emphasizes hands-on, student-centered learning that connects classroom knowledge with practice through labs and coding challenges. Her research focuses on adversarial machine learning, particularly the detection of data poisoning attacks in image classification systems. She is the developer of DynaDetect 2.0, a hybrid framework that combines convolutional neural networks with distance metrics to identify malicious training data. Building on this work, she is advancing DynaDetect-WM, which explores the use of poisoned data as a watermarking strategy for AI content protection. Her current projects also integrate HPC resources and vision–language models to improve both detection accuracy and runtime performance. Dr. Perry is committed to broadening participation in computing. She mentors diverse student teams on senior projects that bridge technical computing with creative fields such as CAD and visual storytelling. Through her research and teaching, she prepares students for impactful careers in AI, cybersecurity, and HPC-enabled applications.
Degrees Earned
Ph.D. in Computer Science, 2025 – University of Mississippi
Research Areas
Computer Science; Data Science; Machine Learning/AI
Research Interests
My research centers on adversarial machine learning, specifically on the detection and mitigation of data poisoning attacks in image classification systems. I am the developer of DynaDetect 2.0, a framework that combines convolutional neural networks with advanced distance metrics to identify malicious training data. This work has demonstrated the importance of statistical and semantic reasoning in ensuring the robustness of machine learning pipelines. Building upon this foundation, I am currently advancing DynaDetect-WM, which explores the use of poisoned data as a watermarking mechanism for content protection in AI models. This emerging direction is particularly relevant to the broader scientific community, where ensuring the integrity of datasets and tracing the origins of AI-generated content is becoming increasingly critical. The integration of HPC resources and vision–language models into both DynaDetect2.0 and DynaDetect-WM strengthens scalability, improves detection accuracy, and expands applicability across scientific domains. This line of work naturally intersects with the interests of project leaders in trustworthy AI, secure data pipelines, and NAIRR-supported large-scale computing projects. By bringing expertise in adversarial resilience and dataset integrity, I hope to contribute meaningfully to collaborative research while offering my students valuable opportunities to engage in the practice of sustainable, responsible AI.
Topical Areas
Artificial Intelligence and Intelligent Systems; Computer Science; Other Computer and Information Sciences
Research Synergy
My field of study, Computer Science, aligns with SRP’s goals through its focus on artificial intelligence, cybersecurity, and high performance computing. My research in adversarial machine learning, particularly the detection of data poisoning attacks, offers an intersection with project leaders working on trustworthy AI, secure data pipelines, and large-scale computational projects. Through DynaDetect2.0, I have developed methods that combine convolutional neural networks with distance metrics to detect poisoned training data. Building on this, DynaDetect-WM extends the framework by using poisoned data as a watermarking strategy for content protection. These approaches complement research across domains that rely on large-scale datasets, from biomedical imaging to natural language processing. With the computational power of NAIRR resources, I aim to test and adapt these methods in new scientific contexts. Engaging with project leaders allows me to contribute expertise in adversarial resilience and dataset integrity, while gaining exposure to applications beyond my current scope. Importantly, this collaboration also creates space for my students to participate in meaningful tasks such as dataset preparation, exploratory runs of watermarking techniques, and collaborative evaluation of results. In this way, our work not only strengthens the robustness of AI pipelines but also advances SRP’s mission of building sustainable, inclusive research partnerships.
Motivation
As a new faculty member at Austin Peay State University, I am eager to grow both as a researcher and as a mentor. Many of my students are first-generation college students who have not had the opportunity to participate in national research collaborations. The Sustainable Research Pathways program offers a way to open those doors, connecting them with meaningful projects while allowing me to learn from experienced project leaders in the NAIRR community. My research focuses on adversarial machine learning and secure data pipelines, but I see this program as more than a chance to extend my own work. It is also a way to strengthen my ability to guide students, especially those from diverse backgrounds, into challenging and rewarding areas of computing. I am excited about the opportunity to collaborate with project leaders whose expertise complements my own, and to bring back new methods, tools, and experiences that will enrich my teaching and research program at APSU. Ultimately, I hope to gain lasting partnerships and practical experience in integrating my students into cutting-edge AI and HPC research. As a new faculty member, being part of a supportive, inclusive community like SRP would help me grow while ensuring my students see themselves as contributors to the future of science.
Supervising Students Plan
I am including Derrick Goralewski, an undergraduate student from my CSCI 1010: Introduction to Programming I course, as the student on my SRP team. Derrick has distinguished himself as an engaged and motivated learner who enjoys programming and has expressed strong interest in advancing in the field of computer science. My plan is to meet with Derrick on a weekly basis to establish goals, provide technical guidance, and review his progress. His responsibilities will include preparing datasets, running small-scale HPC experiments, and documenting results. I will also introduce Derrick to DynaDetect-WM, where he will gain exposure to the concept of using poisoned data as a watermarking strategy for AI content protection. This will allow him to connect his foundational programming skills with emerging research challenges in trustworthy AI. To support his professional growth, Derrick will present his progress during meetings, receive feedback on written work, and practice communicating research findings. My goal is to scaffold his growth from a novice programmer to a confident contributor, ensuring he leaves SRP with meaningful research experience, stronger technical skills, and a deeper sense of belonging in the research community.
Student Merit
I have selected Derrick Goralewski, an undergraduate student currently enrolled in my CSCI 1010: Introduction to Programming I course, to join my SRP team. Derrick has already demonstrated persistence, curiosity, and enthusiasm for programming. He actively engages in class discussions, approaches assignments thoughtfully, and shows a willingness to tackle challenges until he reaches a solution. Although early in his academic career, Derrick has expressed a clear interest in expanding beyond coursework into research. He enjoys programming not just as a requirement but as a field he wants to pursue more deeply, and he has taken initiative to seek opportunities that will allow him to grow. I believe that SRP will provide him with the structured mentorship, exposure to HPC resources, and collaborative environment needed to channel his enthusiasm into productive research experience. Derrick’s preparedness, combined with his eagerness to learn, makes him an excellent candidate for SRP. With proper mentoring, I am confident he will contribute meaningfully to the project while gaining the technical and professional skills that will support his long-term academic and career goals.
Lightning Talk Title
Building Trustworthy AI: My Journey in Cybersecurity
Keywords
Trustworthy AI, Cybersecurity, Data Poisoning, HPC, Watermarking