Recent updates

News

  1. MVPISplat, our defense against attacks on 3D Gaussian Splatting, was accepted to NeurIPS 2026. See publicationsNew

  2. Serving as a reviewer for NeurIPS 2026 and ICWSM 2027.

  3. Our paper on adversarial defense in Gaussian Splatting via 3D Gaussian filtering was accepted to the CVPR 2026 SPAR-3D Workshop.

  4. Our work on detecting stealthy OpenWPM web bots was accepted to IEEE Transactions on Dependable and Secure Computing (TDSC).

  5. Started my PhD in Computer Science at UT Arlington, advised by Dr. Faysal Hossain Shezan.

  6. Wrapped up two great years as a Senior Software Engineer at Samsung Research Bangladesh.

What I work on

Research focus

My research sits at the intersection of computer vision, multimodal learning, and security. I want to understand how modern AI systems, from 3D scene representations to vision-language and generative audio models, can be manipulated, and to build defenses that keep them reliable in the real world.

Security of 3D Vision

Adversarial attacks and defenses for 3D and 4D Gaussian Splatting, such as catching poisoned training views through multi-view photometric inconsistency.

3DGS4DGSAdversarial robustness

Multimodal LLMs

Robustness and security of vision-language models: how adversarial inputs can steer their behavior, and how to defend against it.

VLMsLLMsTrustworthy AI

Generative Audio & Content Protection

Protecting creators' music from unauthorized exploitation by generative AI models.

Audio GenAIData protection

Ubiquitous Computing & Web Security

Applied ML for mobile health and the web, from smartphone-based dehydration detection to behavioral bot detection.

Mobile sensingBot detection

Where I've been

Affiliations