Md Mahbubur Rahman

Md Mahbubur Rahman

PhD Student at Computer Science

Iowa State University

Biography

Md Mahbubur Rahman is currently pursuing a Ph.D. in Computer Science with a focus on AI for Code, Software Engineering, and Causal Deep Learning at Iowa State University, anticipated to be completed by May 2025. His research is concentrated on refining software engineering models through the application of causal deep learning and interpretability methodologies, aiming to enhance prediction accuracy and reliability. His master’s thesis on “Causal Deep Learning for Vulnerability Detection” underscores his commitment to advancing the field.

In his role as a Graduate Research Assistant at Iowa State University, Rahman spearheaded the CausalVul project, which applies do-calculus-based causal learning algorithms to improve transformer-based models for code vulnerability. His industry experience includes a tenure as a Full Stack Software Engineer at crowd-realty.com in Tokyo, where he led the development of REST APIs and data analysis initiatives, and as a Software Engineer at BJIT Inc., where he developed a Bluetooth service for Android and a sentiment-analysis chatbot. Rahman’s expertise bridges the gap between advanced research and practical application in software engineering and AI.

Interests
  • Causal Deep Learning
  • Vulnerability Detection
  • Software Engineering
Education
  • PhD Student in Computer Science, 2025(Expected)

    Iowa State University

  • MS in computer Science, 2023

    Iowa State University

  • BSc in Computer Science and Engineering, 2018

    Jahangirnagar University

Experience

 
 
 
 
 
Iowa State University
Graduate Research Assistant
May 2022 – Present Ames, IA, USA
 
 
 
 
 
Iowa State University
Graduate Teaching Assistant
August 2021 – Present Ames, IA, USA
 
 
 
 
 
Crowd Realty, Inc.
Full Stack Software Engineer
Crowd Realty, Inc.
November 2019 – July 2021 Tokyo, Japan
 
 
 
 
 
BJIT Inc
Software Engineer
BJIT Inc
November 2018 – October 2021 Tokyo, Japan

Achievements

Competitive Programming Excellence
  • 1st place recipient, Inter Dept. Programming Contest, Jahangirnagar University, Bangladesh, 2017.
  • 2nd place recipient, Information Technology Professionals Exam(ITEE FE), Bangladesh, Oct. 2018.
  • 3rd place recipient, National Collegiate Programing Contest, Bangladesh, 2017.
  • 3rd place recipient, 9th ICT FEST, IUT, Bangladesh, 2017.
  • 3rd place recipient, National Programming Camp, Bangladesh, 2016.
  • 4th place recipient, Inter University Programming Contest, NSU, Bangladesh, 2017.
  • 5th place recipient, CSE Day, BUET, Bangladesh, 2016.
  • 12th place recipient, ACM-ICPC Asia Dhaka(Bangladesh) Regional Contest, Bangladesh, 2016.
  • 50th place recipient, Google Kickstart - Round F, 2017.
Online Problem Solving Achievements
  • Codeforces - Max rating 2021 (Top 1%), Username: mahbubcseju.
  • Codechef - Max rating 2309 (Top 1%), Username: mahbubcseju.
  • Hackerrank - Max rating 2345 (Top 1%), 13 gold medals, 11 silver medals, 6 bronze medals, Username: mahbubcseju

Projects

*
CausalVul
Leveraged do-calculus-based causal learning algorithms on top of existing transformer-based vulnerability detection models to systematically eliminate reliance on spurious features, enhancing causal-based predictions.
CausalVul
Empirical Study on DL Models for Code Vulnerability
Surveyed and reproduced 9 state-of-the-art DL models on the Devign and MSR datasets. Investigated model capabilities, training data effects, and interpretability, revealing key insights to improve model robustness and understandability.
Empirical Study on DL Models for Code Vulnerability
Slice Level Vulnerability Detection
Transformer-based models for detecting software vulnerabilities are limited by their token input size, potentially missing crucial data. This project introduces a slicing method that focuses on relevant program points to improve detection accuracy, yielding better performance metrics compared to traditional function-based approaches.
Slice Level Vulnerability Detection
Transformer Explainability
Introduced an explanation method for the transformer-based CodeBERT, which factors in the information flow across layers using Markov chains and integrated gradients for better insight into source code vulnerability predictions.
Transformer Explainability

Publications

(2023). Towards Causal Deep Learning for Vulnerability Detection. In proceedings of the 46th International Conference on Software Engineering, ICSE ’24, Lisbon, Portugal, April 2024.

PDF Cite Code Dataset Project Source Document

(2023). An Empirical Study of Deep Learning Models for Vulnerability Detection. In Proceedings of the 45th International Conference on Software Engineering, ICSE ’23, Melbourne, Australia, May 2023..

PDF Cite Code Dataset Project Source Document

(2023). Modeling Traffic Congestion in Developing Countries Using Google Maps Data. In proceedings of the Future of Information and Communication Conference, 2021, (pp. 513-31). Springer, Cham..

PDF Cite Code Dataset Source Document