Abstract
The rise in educational certificate fraud in India has created significant verification challenges for institutions and employers, as existing manual validation methods are slow, inconsistent, and prone to human error. This paper presents a prototype of a web-based system designed for automated certificate authentication using a multi-modal artificial intelligence pipeline supported by lightweight blockchain hashing. The system integrates Tesseract OCR to extract textual information from uploaded documents and employs a CNN-based model (ResNet) to detect signs of forgery in certificate images. Additionally, each certificate's hash is stored in a shared ledger to ensure tamper-evident verification. Experiments conducted on a dataset of 200 authentic and forged certificates achieved an accuracy of 78% with an average processing time of $3-5$ seconds. The results highlight the potential of AI-driven verification; however, broader datasets and institutional adoption are essential for real-world deployment and improved reliability.