Mohd. Farhan Israk Soumik

Graduate Researcher, Computer Science
Southern Illinois University Carbondale

I am an M.Sc student in Computer Science at Southern Illinois University Carbondale, where I work with Dr. Abdur R. Shahid in the SHIELD Lab. My research studies the security, privacy, and forensic risks that arise when people use consumer generative AI: writing assistants whose one-tap rewrites can both hide emotional signals and help misinformation slip past detectors, and on-device image editors whose edits are hard to localize after the fact.

Before SIUC, I was a research assistant at the Center for Computational and Data Sciences, Independent University, Bangladesh, and earned my B.Sc. in Computer Science & Engineering from RUET. My work has appeared at EMNLP(Main), ACL(IND), ICTAI, ICMLA, and AAAI Workshop on Privacy-Preserving AI(PPAI).

Interests Privacy-preserving ML · Generative AI forensics · LLM & VLM security · Federated learning · Multimodal learning

  • SIU Carbondale
  • CVEX, The Arch
  • RUET
  • IUB
Portrait of Mohd. Farhan Israk Soumik

News

Older news
  • Presented privacy-preserving multimodal stress detection at ICMLA 2025 in Boca Raton, Florida.
  • Joined the SHIELD Lab at SIUC as a Graduate Research Assistant.
  • Received a travel grant to present at the AAAI PPAI-25 workshop.
  • Presented audio-editing defenses against LLM emotion inference at AAAI PPAI-25.
  • Began graduate study in Computer Science at Southern Illinois University Carbondale.

Publications

See also Google Scholar.

  1. EMNLP 2026

    Edits in the Wild: A Multilingual Benchmark for Localizing On-Device Generative Image Edits on Consumer Platforms

    MFI Soumik, AA Fime, WKM Mithsara, AB Aryal, Hoda Joudi Haghighi, Bibek Koirala, Ahmed Imteaj, AR Shahid

    Conference on Empirical Methods in Natural Language Processing (Main), 2026. To appear

  2. ICTAI 2026

    When Quantum Entanglement Meets Transformer Attention: A Controlled Study of Quantum Attention Heads

    MFI Soumik, Hussein Zangoti, Ahmed Imteaj, AR Shahid.

    38th IEEE International Conference on Tools with Artificial Intelligence, 2026. To appear

  3. ACL 2026

    No Innocence in Styling: Discovery of Privacy Protection Capabilities and Security Risks in Consumer Generative AI Writing Assistants

    MFI Soumik, SM Hasan, MWW Kankanamge, A Imteaj, AR Shahid

    Annual Meeting of the Association for Computational Linguistics (Industry Track), 2026.

  4. ICMLA 2025

    Privacy-Preserving Multimodal Stress Detection from Wearables with Attention Fusion and Federated Learning

    MFI Soumik, E Maghsoudlou Nima, H Zangoti, A Imteaj, AR Shahid

    24th IEEE International Conference on Machine Learning and Applications, 2025.

  5. AAAI-W 2025

    Audio Editing Features as User-Centric Privacy Defenses Against Large Language Model (LLM) Based Emotion Inference Attacks

    MFI Soumik, WKM Mithsara, AR Shahid, A Imteaj

    6th AAAI Workshop on Privacy-Preserving Artificial Intelligence (PPAI-25), 2025. Travel grant

  6. arXiv 2025

    Evaluating Apple Intelligence's Writing Tools for Privacy Against Large Language Model-Based Inference Attacks: Insights from Early Datasets

    MFI Soumik, SM Hasan, AR Shahid

    arXiv preprint arXiv:2506.03870, 2025.

  7. TENSYMP 2020

    Brain Tumor Classification with Inception Network Based Deep Learning Model Using Transfer Learning

    MFI Soumik, MA Hossain

    IEEE Region 10 Symposium (TENSYMP), 2020.

Research Experience

Education

Teaching

Awards & Service

Beyond Research

I got into computer science by watching my older brother write code at home. Seeing a few lines of text turn into a working program felt like quiet, everyday magic, and I have wanted to understand how it works ever since.

I'm from Bangladesh and moved to Carbondale in 2024. I speak Bangla and English, and some German and Hindi. Before graduate school I taught GRE prep for a year.

Standing beside my poster at ACL 2026 in San Diego
With my poster at ACL 2026, San Diego