📄 Sciences Methods and Technologies
International Journal (SciMeTech)

Machine learning (ML) is based on a core concept in artificial intelligence (AI) which is data training that makes ML models understand patterns and relationships. Since ML systems heavily rely on the integrity of training data, it's essential to protect the datasets from tampering and data poisoning attacks, which can significantly degrade model performance. In this paper, we propose a lightweight and practical framework that leverages blockchain technology to ensure the integrity of training datasets. The proposed approach computes a cryptographic hash of the dataset and stores it on a blockchain-like structure, enabling verification prior to model training. We implement a prototype using Python and evaluate the system under normal and tampered data scenarios. Experimental results demonstrate that the proposed method effectively detects dataset modifications with negligible overhead. This work provides a simple yet effective solution for enhancing trust in ML pipelines.