Research Vault
Optimizing Post-Quantum LWE Encryption for Large Image Data
B.Sc. Thesis, SUST(In Progress) • 2025-26
A comprehensive hybrid cryptographic preprocessing and acceleration framework integrating image compression techniques (RLE, Huffman coding, YCbCr with chroma subsampling) with Strassen's matrix multiplication for LWE-based encryption. Achieved 41.19% average data volume reduction while maintaining lossless compression (PSNR = ∞) and full post-quantum security guarantees.
BanglaNoise-10: A Diverse Audio Dataset for Machine Learning on Urban Environmental Noise in Bangladesh
Data in Brief (Under Review) • 2025
First nationwide South Asian environmental noise dataset containing 5,035 labeled audio clips across 10 urban noise categories. Achieved 98% accuracy with Whisper-Base, 97% with Wav2Vec2-Base. Dataset includes bike, bus, car, CNG auto-rickshaw, construction noise, protest, siren, traffic jam, train, and truck sounds collected across Bangladesh.