I study how intelligent systems can become more reliable, culturally aware, and accountable. My current work combines technical deepfake detection with data-centric evaluation, platform research, and the social realities of AI-generated media in the Global South.
Working with Dr. Dilrukshi Gamage, Dr. Savindhu Herath, and Prof. Yash Raj Shrestha on a region-wide study of deepfake detection in South Asia. The work examines the methods, datasets, modalities, and evaluation practices represented in the literature, with particular attention to regional languages, identities, media practices, and underrepresented populations.
Period: November 2025–Present
Investigating how deepfake-related discourse changes as it moves from Reddit into public Telegram channels. The study combines computational analysis with qualitative coding to examine creation, sharing, detection, concern, abuse, and media-literacy narratives across languages and communities.
Period: 2026–Present
Synthesising evidence from more than 70 peer-reviewed studies to understand how mammogram preprocessing affects downstream breast-cancer detection. The work maps common enhancement, denoising, segmentation, and normalisation techniques while identifying gaps in adaptive and clinically robust preprocessing pipelines.
Supervised by Dr. Omer Bin Saeed and Dr. Shahab Uddin Ansari.
Status: Manuscript in writeup stage
Designed and developed an end-to-end framework for verifying images, videos, and audio. AVIA integrates ConvNeXt for image forensics, GenConViT for video analysis, and a CNN–BiLSTM architecture for synthetic-speech detection.
The system was trained and evaluated on 12 benchmark datasets containing more than 5,000 samples, with specialised pipelines for GAN-generated and diffusion-generated media. This project formed my undergraduate thesis at GIKI under the supervision of Dr. Khurram Khan Jadoon.
Period: February 2024–May 2025
Examined why developing countries are particularly vulnerable to the political and social harms of deepfakes. The study considers disinformation, electoral interference, reputational harm, fraud, legal gaps, and declining trust in media, and argues for a response that combines technical detection, public awareness, governance, and interdisciplinary collaboration.
Published: 23 September 2024 · Cited 100+ times
Developed and evaluated a hybrid RFID–PIN authentication mechanism designed to improve access control through two independent verification layers. The work was presented through MTME 2025 and published in Materials Proceedings.
Published: 2025
