Md Jahirul Islam

Ph.D. student in Computer Science, Kennesaw State University

Md Jahirul Islam

I study how deep learning models fail under attack—backdoors, data poisoning, and physically realizable adversarial patches—and how to defend remote-sensing models and large language models against them.

View CV
Lab
ASTAI Lab, advised by Dr. Kazi Aminul Islam
Focus
AI security and privacy in remote sensing and large language models
Portrait of Md Jahirul Islam in graduation regalia at Kennesaw State University

About

I'm a Ph.D. student in Computer Science at Kennesaw State University and a Graduate Research Assistant in the ASTAI Lab, advised by Dr. Kazi Aminul Islam. My research is in AI security: physically realizable backdoor attacks and clean-label poisoning against remote-sensing models, the vulnerability of large language models, and defenses that identify and purify adversarial inputs. My first-author paper on robust physically realizable backdoor attacks appeared at IEEE TPS 2025, and our work on adversarial patch purification with spectral and diffusion models is under review at AAAI 2027.

Before the Ph.D., I completed a master's degree in Computer Science (AI concentration) at KSU, where I worked in the HIX Lab on real-time cybersickness detection and mitigation for standalone VR headsets—work supported by the U.S. Army Research Laboratory and published at IEEE VR 2024 and 2025. Earlier, I built production computer-vision systems as a Machine Learning Engineer at AI Samurai Japan, and I earned my B.Sc. in Computer Science and Engineering from Chittagong University of Engineering & Technology, Bangladesh.

I also review for IEEE venues, including the IEEE Internet of Things Journal, and I'm open to collaborations on adversarial robustness, model security, and trustworthy AI.

Experience

Research

  1. Jan 2025 – present

    Graduate Research Assistant

    ASTAI Lab, Kennesaw State University

    • Investigate AI security and physically realizable adversarial attacks on AI models for remote-sensing applications.
    • Develop clean-label adversarial attacks for remote-sensing AI models.
    • Investigate the vulnerability of large language models.
    • Develop defense mechanisms that improve model robustness against adversarial attacks.
  2. Aug 2023 – Dec 2024

    Graduate Research Assistant

    HIX Lab, Kennesaw State University

    • Developed cybersickness-prediction models from physiological data (heart rate, electrodermal activity) combined with head- and eye-tracking data.
    • Built deep learning models with RNN, GRU, LSTM, CNN–LSTM, ResNet, and NHITS architectures. Code
    • Deployed models on the Meta Quest 3 with Unity for real-time cybersickness prediction on its Snapdragon processor.
    • Developed a framework for real-time cybersickness detection and mitigation using cybersickness time-series data.
    • Worked with the HP Reverb G2 Omnicept, EmotiBit, ESP32, and virtual- and mixed-reality technologies.

Industry

  1. Jan 2023 – Jul 2023

    Machine Learning Engineer, Computer Vision

    AI Samurai Japan

    • Developed a car image-segmentation model with YOLOv6 and took it to production with Flask, Docker, GCP, and GitHub Actions.
    • Built an anomaly-detection model for potato classification using PatchCore and PaDiM, with DINO and SAM for automatic segmentation.
    • Fine-tuned YOLOv8, YOLOX, RTMDet, and other state-of-the-art models on several image datasets.
    • Developed SMPL-model functionality and 3D and depth-map generation in Blender.

Publications

Google Scholar profile

Peer-reviewed conference papers

  1. 2026

    MediBox+: A Smart Dispenser with Mobile App for Enhancing Medication Adherence

    S. Sharmin, M. J. Islam (equal contribution)

    ICSADL 2026

    Paper

  2. 2025

    Robust Physically Realizable Backdoor Attack

    M. J. Islam, K. A. Islam

    IEEE TPS 2025

    Paper

  3. 2025

    Comparative Analysis of Defense Approaches against Adversarial Attacks on Remote Sensing Applications

    S. Ravikumar, M. J. Islam

    IEEE ICMLA 2025

    Paper

  4. 2025

    Towards Optimized Real-time Cybersickness Detection Framework Using Deep Learning for Standalone Virtual Reality Headsets

    M. J. Islam, R. Islam

    IEEE VR 2025

  5. 2024

    Towards Optimized Cybersickness Prediction for Computationally Constrained Standalone Virtual Reality Devices

    M. J. Islam, R. Islam

    IEEE VR 2024

    Paper

  6. 2021

    Smart Water Quality Monitoring and Controlling System

    M. J. Islam, Asaduzzaman

    Proceedings of the 5th International Conference on Electrical Information and Communication Technology (EICT), Khulna, Bangladesh, December 2021

    Paper

Under review

  1. 2026

    SpectraDiff: Adversarial Patch Identification and Purification through Spectral and Diffusion Models

    H. Perry, M. J. Islam, K. A. Islam

    Submitted to AAAI 2027

Thesis

  1. 2024

    Real-Time Feedback-Driven Framework for Automated Cybersickness Mitigation

    M. J. Islam

    Master's thesis, Kennesaw State University

    Thesis

Other scholarly outputs

  1. 2025

    Evaluating Transferability of Adversarial Attacks Between Models of Different Architectures

    S. Noboudem, R. K. Rogannagari, M. J. Islam

    Kennesaw State University Digital Commons

    Repository

  2. 2025

    Decentralized Scheduling and Memory Management in a Simulated Multi-kernel OS Environment

    M. J. Islam

    Kennesaw State University Digital Commons

    Repository

  3. 2023

    Sudoku Solver Using Brute-force Algorithm with Backtracking Approach

    M. S. Alam, M. J. Islam, et al.

    Kennesaw State University Digital Commons

    Repository

Projects

Demo videos on YouTube

  • Next Bazar: everyday grocery price forecasting

    2026

    A live dashboard that tracks and forecasts the retail prices of 84 everyday groceries in Bangladesh. A scraper collects the Trading Corporation of Bangladesh's daily market-price sheets (May 2021 to the present); an XGBoost and LightGBM ensemble, benchmarked against Prophet and SARIMAX with walk-forward validation, produces 7-day and 90-day forecasts; and a GitHub Actions job refreshes the data, models, and dashboard every day.

    Python, XGBoost, LightGBM, Prophet, time-series forecasting, React, GitHub Actions

    Live dashboard Code

  • Cybersickness prediction for standalone VR headsets

    2023–2024

    RNN, GRU, LSTM, CNN–LSTM, ResNet, and NHITS models that predict cybersickness from physiological and head- and eye-tracking data, deployed on the Meta Quest 3 with Unity for real-time inference.

    Deep learning, time series, Unity, Meta Quest 3

    Code

  • Brain tumor segmentation

    2024

    A custom 3D U-Net trained on the BRATS dataset for volumetric tumor segmentation.

    3D U-Net, BRATS, medical imaging

    Code

  • Diabetic retinopathy classification

    2024

    Transfer learning and fine-tuning of a pretrained model to classify diabetic retinopathy from retinal images.

    Transfer learning, computer vision

    Code

  • Breast cancer prediction

    2024

    Classification with a support vector machine.

    SVM, machine learning

    Code

  • Optimized hybrid load balancer

    Dynamic network-traffic distribution and performance enhancement with a hybrid load-balancing strategy.

    Networking, systems

    Code

  • Medicine reminder and dispenser

    2023

    A dispenser and companion mobile app that deliver buzzer, LED, and phone reminders together with medication information.

    Arduino, NodeMCU, ESP32, RTC module, Java

    Code Video

  • Mango disease detection

    2021

    An Android application that classifies mango diseases with convolutional neural networks.

    CNN, TensorFlow, Android

    Code

  • Smart car parking system

    2021

    Parking-space discovery, automated parking, fee calculation, and security features.

    Android, Arduino, ESP8266, ESP32, Firebase, Embedded C

    Code

More projects

  • Image preprocessing toolkit

    2023

    Image preprocessing and alignment with OpenCV.

    Code

  • USA universities requirements finder

    2022

    A tool for finding and compiling US university admission requirements, built with Python, Selenium, pandas, and Excel.

    Code

  • Depression test and counseling

    2021

    A self-assessment application with doctor contact information.

    Google Play

  • Sales record inventory

    2020

    An Android application for distributors to manage products, prices, delivery personnel, and payments.

    Video

  • Smart door lock

    2019

    An RFID- and password-operated lock built with Arduino, ESP8266/ESP32, Firebase, Embedded C, and RTOS.

    Code Video

  • Water quality monitoring at Kaptai Navy Camp

    2019

    Sensor-based water quality monitoring deployed in a real-world setting.

    Code

  • Electricity bill management system

    2019

    A billing and database-management website built with HTML, CSS, SQL, and PHP.

    Code

  • Plover

    2019

    An online chat application built with Java, Gradle, Firebase, and XML, published on Google Play.

    Code

  • Arduino code for robotics projects

    A collection of Arduino sketches from robotics projects.

    Code

Skills

Core
Adversarial attack generation, vulnerability testing, image processing, 3D modeling, data analysis, data and model distillation, cloud infrastructure management, AI model deployment on servers, ethical AI practices
Languages
Python, C, C++, C#, Java, Embedded C
Machine learning and computer vision
PyTorch, TensorFlow, Keras, scikit-learn, OpenCV, CNNs, LSTMs, GANs, diffusion models, LLMs, VLMs
Databases
Firebase, MySQL, PostgreSQL, SQLite, NoSQL
Virtual reality and embedded systems
Unity, Blender, mobile application development, Internet of Things
Cloud and development tools
AWS, GCP, Docker, Git, GitHub, Unix
Software and GPU tools
Flask, REST APIs, Selenium, NVIDIA CUDA programming, MCP agents, LaTeX

Education

  1. Jan 2025 – present

    Ph.D. in Computer Science

    Kennesaw State University, Georgia, USA

    In progress; comprehensive examination passed. GPA 4.00/4.00.
    Research: AI security and privacy in remote sensing and large language models.

  2. Dec 2024

    Master's in Computer Science, AI concentration

    Kennesaw State University, Georgia, USA

    GPA 4.00/4.00.
    Thesis: Real-Time Feedback-Driven Framework for Automated Cybersickness Mitigation.

  3. Aug 2021

    B.Sc. in Computer Science and Engineering

    Chittagong University of Engineering & Technology, Bangladesh

    GPA 3.12/4.00.
    Thesis: Smart Water Quality Monitoring and Controlling System.

Service and recognition

Professional service

Awards

  • 2025IEEE Student Travel Grant Award
  • 2025CCSE Graduate Student Research Travel Award
  • 2025Graduate College Research Travel Award
  • 2024CCSE Graduate Student Research Travel Award
  • 2016First Runner-Up, Intra-University Tech Carnival, CUET Students Union

Leadership and extracurricular

  • 2023–2025Organizing Committee, Bangladesh Student and Teachers Association at KSU
  • 2019Instructor, workshop on basic electronics and IoT technology
  • 2018–2019Joint General Secretary, Robo Mechatronics Association, CUET
  • 2018–2019Organizing Committee, CUET Computer Club
  • 2018Volunteer, Inter-University Robotic Competition, RMA, Bangladesh
  • 2017Organizing Committee, Inter-University Programming Contest at CUET

Certifications and training

Contact

Send me a message about collaborations, reviewing, research questions, or anything else. I usually reply within a few days.