The Center of Excellence in Cybersecurity Research, Education and Outreach at North Carolina Agricultural and Technical State University has announced a new lecture in its month-long series in recognition of Cybersecurity Awareness Month.
The speaker for "Career Opportunities as a Scientist/Engineer in Support of National Defense in the Department of War" is a civilian scientist at the Naval Surface Warfare Center (NSWC) Crane Division, and will discuss career paths, internships, and scholarships available to students.
Dr. Bryan Woosley will highlight programs such as the Office of Naval Research's enterprise internship, the STEM Student Employment Program, and the Science, Mathematics, and Research Transformation (SMART) Scholarship.
As university liaison to the chief technology officer and site coordinator for the Naval Engineering Education Consortium, Dr. Woosley oversees multiple STEM initiatives and facilitates research collaborations.
He will speak with undergraduates in the College of Science and Technology and the College of Engineering about opportunities ranging from internships to scholarships that support degrees and connect students to civilian roles in the Department of Defense.
The event is scheduled for Thursday, October 29, at NC A&T.
Earlier in the week, the Center hosted a talk addressing a critical issue in computing: the risks when intelligent agents act on unintended instructions.
Modern agents can browse, read files, and take actions, but may not distinguish between legitimate commands and malicious content.
Attackers can exploit this by embedding hidden instructions in web pages, PDFs, or support tickets.
Bedeabasi John, a PhD researcher at the Center, explained why system prompts alone are insufficient and discussed how sandboxing, read-only defaults, and human approval can help prevent intelligent agents from being misused.
His research focuses on privacy-preserving machine learning and countermeasures against privacy attacks in deployed AI systems, with current work examining persistent privacy risks in agentic AI.
On October 22, Dr. Mohammad Arif Ul Alam, associate professor of computer systems technology at NC A&T, discussed how voice data can reveal personal and health information, making it both a valuable tool for healthcare AI and a privacy risk.
Dr. Alam addressed threats such as profiling, impersonation, and voice cloning, and highlight research opportunities in privacy-preserving AI.
Before joining A&T, Dr. Alam held positions at UMass Lowell, the MIT-IBM Watson AI Lab, and the National Institutes of Health.
On October 6, the Center launched its 2026 speaker series with Tia Pope, an assistant professor of computer science at NC A&T.
Pope emphasized the importance of including intelligent agents in security threat models, noting that agents can read, choose tools, and act, which increases risk.
She outlined three ways agents can be compromised: receiving malicious goals, following hidden instructions, or exceeding their intended scope.
Real-world examples included a coding agent deleting production data, attackers gaining admin control of a school network in minutes, and malicious add-ons in agent marketplaces.
Pope stressed the need for clear agent identity, defined scope, accountable ownership, enforced controls, and continuous monitoring.
She concluded by asking: What would you allow an agent to do, and how would you monitor it?
Recently, Pope, then a third-year PhD candidate at N.C. A&T University delivered a lightning talk at All Things Open AI in Durham.
Her research centers on evaluating and developing artificial intelligence tools for protein design.
Pope also contributes to open-source initiatives to broaden access to advanced protein-design tools.
Pope described how transformer-based models, such as ProtGPT2 and ESM, are transforming protein sequence analysis by enabling detailed embeddings and advanced function prediction. Her presentation provided an overview of pre-trained, open-source transformer models that generate protein embeddings and their application in classification tasks.
She also demonstrated methods for tokenizing sequences, extracting embeddings, and constructing machine-learning pipelines for protein function annotation using Gene Ontology (GO) or Enzyme Commission (EC) numbers.
The presentation additionally addressed topics such as dark proteins and AlphaFold.
