Vulnerability Detection with Interprocedural Context in Multiple Languages: Assessing Effectiveness and Cost of Modern LLMs
2026
Research
/focus
I study how large language models can support secure software engineering, especially vulnerability detection and automated repair. My work focuses on practical methods for reasoning about production codebases, interprocedural context, multiple programming languages, and the cost of using modern LLMs in vulnerability analysis workflows.
/background
I am a Ph.D. student in Computer Science at North Carolina State University, advised by Wesley K. G. Assunção. My M.S. research at the Federal University of Alagoas focused on large language models for automated vulnerability repair across multiple programming languages. My undergraduate work produced a multi-language dataset of open-source software vulnerabilities and their fixes.
/topics
- LLM-based vulnerability detection with interprocedural code context.
- Cross-language vulnerability repair and automated program repair.
- Secure development workflows for real-world software systems.
- Automated software testing with generative AI and browser automation.
- Usability, assistive technology, and applied software systems.
/papers
Beyond Code Explanations: A Ray of Hope for Cross-Language Vulnerability Repair
2025
Turning Manual Tasks into Actions: Assessing the Effectiveness of Gemini-generated Selenium Tests
2025
/earlier
Before my current LLM security research, I worked on applied software systems for judicial automation, scientific computing, and assistive technology.
EASY/UFAL
Developed CliqueJunte, Cassandra, and Laura. These applied systems support document attachment in PJe workflows, preparation of RPV/PRC requisitions through Jurisdição Delegada, and AI-assisted triage of medical expert reports.
LCCV/UFAL
Developed SimWear as part of computational research on oil-well casing wear prediction, supporting tools that use historical well-trajectory data, statistical analysis, and wear models to estimate mechanical degradation.
TOCA/UFAL
Developed TATU, a mobile app for visitors who are blind, have low vision, or are deaf or hard of hearing. The project combined IoT, audio description, and Libras so cultural attractions could be experienced with more autonomy.
/software
TATU - Tecnologia Assistiva Aplicada ao Turismo
SGA - Sistema de Gerência de Atrações Turísticas
TutoMath
/profiles
Find my research on Google Scholar, ORCID, and Lattes.