Integrating LLM with Blockchain and IPFS to Enhance Academic Diploma Integrity
Jean Gilbert Mbula Mboma, Kuti Lusala, Michel Matalatala, Obed Tshimanga Tshipata, Pierre Sedi Nzakuna, Dave Tshimbalanga Kazumba · 2024
This paper presents an innovative solution to counter academic document fraud. It consists in combining a Large Language Model (LLM), specifically the Bidirectional Encoder Representations from Transformers (BERT), with blockchain and Interplanetary File System (IPFS) to pre-validate academic documents before their certification. This approach addresses the vulnerabilities inherent in current systems, particularly the over-reliance on the trustworthiness of system users issuing documents. The results demonstrate that our approach outperforms simple IPFS-blockchain-based solutions in detecting document alterations. We achieved an accuracy of 99.3% with a slight increase of 1 second in latency per document. Although the study has been carried out with high school diplomas delivered in the Democratic Republic of the Congo (DRC), the proposed solution has broad applicability. It can be globally adapted for various academic document verification purposes.