Tackling Cybersecurity Attacks and Software Vulnerabilties using Machine Learning
Abstract
The overarching research theme of this project is a machine learning driven approach to address important research problems in cybersecurity. The project has two fundamental research goals. The first is using machine and deep learning techniques to tackle deceptive attacks and software vulnerabilities. Deceptive attacks such as phishing, spearphishing, job scams, Business Email Compromise and fake news are proliferating and causing significant damage and loss of trust. Another serious security concern is intrusions because of bugs in software. This project will investigate machine learning and natural language processing techniques for these two important problems. We plan to make fundamental advances on linguistic cues, data augmentation techniques, and deep learning models for domain-independent deception. For software vulnerabilities we plan to advance the state of the art in verifying correctness and locating bugs through static analysis, teacher student learning, abstractions, multimodal models and develop quality datasets. The second goal, which is complementary to the first, is to devise machine learning techniques that can efficiently scale to work on large data. Machine learning techniques need to operate on large-scale data due to increasing sizes of the cybersecurity data sets being collected. Scaling these techniques to work on large data sets in an efficient way is critical for taking effective cybersecurity decisions in a timely fashion. The project will investigate scalable machine learning techniques by devising efficient distributed algorithms that can substantially speed up these techniques while guaranteeing a similar quality of performance as in traditional (centralized) settings. The project will lead to fundamental advances in distributed algorithms for machine learning. It will develop a set of machine learning techniques, algorithms, and efficient implementations. It will also test these implementations on real-world data sets that arise in cybersecurity applications. The project will (a) significantly advance the ability to solve important research problems in cybersecurity which are very relevant to the national security functions of DoD (b) enhance the capacity of University of Houston, a Tier 1 minority serving public research institution to participate in DoD research programs and activities; and (c) increase the number of graduates, including underrepresented minorities, in fields of science, technology, engineering, and mathematics (STEM) relevant to the DoD goals.
Document Details
- Document Type
- DoD Grant Award
- Publication Date
- May 24, 2023
- Source ID
- W911NF2310191
Entities
People
- Rakesh M. Verma
Organizations
- Army Contracting Command
- Office of the Secretary of Defense
- University of Houston