I am a researcher in artificial intelligence and cybersecurity at the School of Computing, Mathematics and Physics, University of Portsmouth, working on the Horizon Europe XTRUST-6G project on Zero Trust Architecture for next-generation networks.
I hold an MSc in Cyber Security and Forensics (Distinction) from the University of Portsmouth and a BSc in Computer Science and Engineering from Daffodil International University, Bangladesh. My earlier research applied machine learning to healthcare and agriculture, including early detection of dementia and identification of medicinal plants.
A platform for hands-on cybersecurity teaching and training: Docker-based challenge hosting, OpenVPN team isolation, real-time scoring and interactive penetration-testing environments. In daily production use for lab sessions by multiple universities and departments, including the University of Portsmouth, and expanding across the UK and Europe.
Research and engineering on Zero Trust Architecture for secure next-generation networks: continuous verification of users, devices and services, policy-driven access control and dynamic trust evaluation, as part of a European consortium.
A forensic model for distinguishing real from AI-generated images. DeepGuard combines zero-shot CLIP embeddings with handcrafted forensic features (pixel-level inconsistencies, PRNU sensor patterns) in a lightweight, explainable classifier suited to investigators and researchers.
AI Threat Classifier, ensemble deep-learning classifier for malicious payloads and anomalous network traffic · Django IP Lock, Django middleware limiting login attempts by IP · ThinkAI, text-to-image, PDF and document generation platform · MockTak, community and analytics platform for IELTS learners · BechDia, web and Android marketplace for Bangladesh, published on Google Play · Python Virtual Assistant, speech-driven task assistant.
The Universe is a space flight simulator that runs in a web browser. I designed, built and shipped it on my own. The player lifts off from Kennedy Space Center, reaches orbit, travels to any of 57 real places in space, from the Moon and Europa out to nebulae, black holes and the Andromeda galaxy, then lands on a planet or a moon and walks its surface under that planet's real gravity.
Nothing in the sky is decorative. Every position, size, distance and travel time for light is worked out live from published astronomical data at the current clock time, so the planets are where they really are while you are flying. The project's design constraint was honesty: where the simulation departs from physics, in the wormhole that removes the travel time, the guided landings and the close approaches to distant stars, the game states it on screen, next to the real numbers.
What RAG (retrieval-augmented generation) actually does, why LLMs hallucinate without it, how chunking and embeddings work, and where RAG quietly fails.
How AI-generated image detection actually works: camera fingerprints, frequency artefacts, CLIP embeddings, and why your eyes are the least reliable tool.
Overfitting explained in plain words: overfitting vs underfitting, data splits, why rising validation loss is the tell, and how regularisation stops it.
A practical Docker security hardening checklist: non-root users, read-only filesystems, dropped capabilities, no Docker socket, digests and limits.
How machine learning intrusion detection works: flow features, random and isolation forests, dataset pitfalls, class imbalance, drift and SIEM deployment.