cv

Curriculum Vitae of Jubair Ahmed Nabin, Lecturer & Researcher in CSE.

Contact Information

Name Jubair Ahmed Nabin
Professional Title Lecturer & ML/Cybersecurity Researcher
Email janabin.research@gmail.com
Location Dhaka, Bangladesh
Website https://nabin47.github.io

Professional Summary

Lecturer and researcher in Computer Science & Engineering at IUBAT, Dhaka. Research focuses on federated learning, intrusion detection systems, explainable AI (SHAP/LIME), adversarial machine learning, and video anomaly detection. Author of 7 peer-reviewed papers at IEEE and Springer venues.

Experience

  • 2023 -

    Dhaka, Bangladesh

    Lecturer, Department of CSE
    International University of Business Agriculture and Technology (IUBAT)
    Teaching undergraduate courses in CSE and mentoring student research projects.
    • Courses: Artificial Intelligence, Data Science, Data Structures, Database Management Systems, System Analysis & Design
    • Mentored 35+ undergraduate students on ML and cybersecurity research projects; two student-led projects advanced to peer-reviewed publication (IEEE PECCII 2026, IEEE ICMI 2026)
  • 2022 -

    Dhaka, Bangladesh

    Researcher — Machine Learning & Cybersecurity
    CUET / IUBAT
    Independent research in federated learning, intrusion detection, adversarial ML, and video anomaly detection.
    • Developed a privacy-preserving CNN-LSTM architecture within a Federated Learning framework for network intrusion detection, achieving over 99.36% accuracy across all seven attack classes on CIC-IDS2017 without centralizing raw data; raised Web Attack F1-score from 0.20 (CNN baseline) to 0.93, outperforming LSTM, CNN+BiLSTM, and CNN+GRU variants
    • Built a novel file-entropy dataset and detection pipeline for crypto-ransomware, developing a custom FUSE-based filesystem to intercept file write/delete operations across 123 file types (30 ransomware strains, 93 benign); trained ML models achieving up to 96% detection accuracy (first-author, Springer LNNS 2025)
    • Evaluated cross-domain adversarial robustness and XAI explanation stability in phishing detection, comparing PGD and Carlini-Wagner attack accuracy alongside SHAP explanation drift across feature-engineered vs. raw-URL input representations (IEEE PECCII 2026)
    • Designed a two-stage CNN + Vision Transformer pipeline for video anomaly detection on the UCF-Crime dataset, reaching 98% accuracy on binary classification and 95% on multi-class categorization; introduced a keyframe extraction algorithm reducing per-video processing time to 20ms

Education

  • 2018 - 2023

    B.Sc.
    Chittagong University of Engineering & Technology (CUET)
    Computer Science and Engineering
    • Machine Learning
    • Data Structures & Algorithms
    • Computer Networks
    • Database Management Systems
    • Artificial Intelligence
    • Software Engineering

Awards

  • 2026
    Research Publication Recognition
    Miyan Publications Reward Ceremony
  • 2022
    5th Position — CUET Talent Hunt IUPC
    CUET

    Competitive programming contest, 2018–2022

  • 2022
    Round 1 Qualifier
    Meta Hacker Cup

    2020 and 2022

  • 2017
    General Board Scholarship
    Board of Intermediate and Secondary Education

    2015 and 2017

Publications

Skills

Programming Languages: Python, C++, SQL, JavaScript
ML / DL Frameworks: PyTorch, TensorFlow, Keras, Scikit-learn
Federated Learning: FL aggregation pipelines, privacy-preserving distributed training
Explainability & Security: SHAP, LIME, PGD, Carlini-Wagner attacks, adversarial robustness
Data & Visualization: Pandas, NumPy, Matplotlib, OpenCV
Tools & Environments: Git, Linux, LaTeX, Jupyter Notebook, VS Code, Google Colab

Projects

  • FL-IDS Framework
    • Evaluated on CIC-IDS2017 across seven attack classes; 99.36% accuracy
    • Web Attack F1-score raised from 0.20 (CNN baseline) to 0.93
  • Crypto-Ransomware Detection Pipeline
    • 30 ransomware families and 93 benign file categories
    • ML classifiers achieving up to 96% detection accuracy
  • Adversarial Phishing Detection Study
    • Compared PGD and Carlini-Wagner attacks across feature-engineered and raw-URL inputs
    • Feature-engineered inputs yield higher adversarial accuracy and lower explanation drift
  • Video Anomaly Detection System
    • 98% binary and 95% multi-class accuracy
    • Keyframe extraction algorithm reduces per-video processing time to 20ms