Page 32 - TECH MAGAZINE CSE

AN
VESHAN

BEST PROJECT SYNOPSIS

SecureAI: Blockchain-Driven Decentralized Data Storage for
Robust Deep Learning

Agastya Todil (01114802720), Dr. Sudha Narang, Ayush Agarwal (03514802720)

Deep learning models and artificial intelligence (AI) advancements have completely
changed a number of industries. But worries about privacy, data integrity, and the
reliability of AI systems continue. To address these issues, this project presents a novel
solution that incorporates blockchain technology into the deep learning application
storage and retrieval process, with a focus on supply chain management (SCM). To create

an immutable ledger, the project uses a custom blockchain class. Each dataset entry is
stored as a block with SHA-256 hashing and timestamping for increased privacy and
robustness. Data reliability is ensured through the integration of quality control
mechanisms. A recurrent neural network model known as the Long Short-Term Memory
(LSTM) model is trained on the stored dataset in order to forecast SCM-related features
over time. The project assesses the effectiveness of the LSTM model on blockchain-stored
datasets in comparison to conventional CSV-stored datasets, in addition to showcasing
the viability of blockchain-based dataset storage. The findings highlight the potential
benefits of blockchain technology in improving the privacy, data integrity, and general
trustworthiness of AI systems. In the future, the architecture will be scaled to
industrystandard datasets and models, and new blockchain features like smart contracts
and ledger functionalities will be investigated.