Research Symposium

BIO

Aakanksha Pathak is a junior at Florida State University pursuing a Bachelor of Science in Computer Science with a minor in Computational Entrepreneurship. She previously earned an Associate of Arts in Computer Science, cum laude, from Tallahassee State College, where she received President’s List recognition and participated in the Honors Program.

Pathak’s academic and research interests focus on the intersection of artificial intelligence, cybersecurity, and emerging technologies. She participated in the MSIPP REU Program at the FAMU-FSU College of Engineering, where she developed a physics-informed Graph Neural Network (GNN) to detect defects in metal additive manufacturing using sparse stereovision data. Her research also explores AI-based vulnerability detection in Windows binaries, applying machine learning techniques such as convolutional neural networks and probabilistic models to analyze executable files without disassembly. She conducts this research under the mentorship of Dr. Sharanya Jayaraman in the Department of Computer Science.

In addition to research, Pathak is active in leadership and innovation initiatives on campus. She is a founding member and Events Coordinator of the Innovation Club at Florida State University, where she helps organize programs that encourage entrepreneurship and technology-driven collaboration among students. She also founded the Code Busters coding club and previously served as Treasurer and Parliamentarian in Student Government. Pathak plans to pursue advanced work in artificial intelligence and cybersecurity, potentially through graduate study and research-focused industry roles.

AI-DRIVEN VULNERABILITY AND MALWARE DETECTION IN WINDOWS

Abstract

Malware and software defects in Windows applications continue to pose serious
security risks, especially since hackers are using techniques to hide or disguise
malicious activity.
This study uses machine learning (ML) and artificial intelligence (AI) techniques to
analyze Windows executable files without disassembly in an effort to improve
malware and vulnerability detection. The project examines whether structural and
behavioral patterns alone can identify serious security risks by viewing each program
as a "black box."

The study utilizes datasets of both known benign and malicious Windows binaries.
Characteristics such as byte-level patterns, entropy values, API calls, and PE-header
details are extracted and then refined through normalization and feature selection.
These characteristics are subsequently employed to train and assess machine
learning and deep-learning models, including Convolutional Neural Networks
(CNNs), Markov Chain-based models, and other types of classifiers, to differentiate
malicious files from those deemed safe.

Initial findings suggest that artificial intelligence models can effectively identify
harmful activities, including hidden or disguised malware, with greater accuracy than
traditional methods that rely on signatures.
The results suggest that analyzing patterns is a useful way to find threats that
traditional methods might miss, so it can be applied to larger or more complex
datasets.
In summary, this research contributes to the development of faster, more scalable,
and more reliable methods for detecting malware.
The results advocate for the implementation of AI-based strategies to enhance
cybersecurity measures and improve the automated evaluation of Windows software.

Keywords: Machine learning, Window binary, Artificial Intelligence