TokenDoc: Source Authentication With a Hybrid Approach of Smart Contract and RNN-Based Models With AES Encryption

This study proposes a hybrid architecture in which duplicates in the dataset are identified using a deep-learning neural network to authenticate the source. The blockchain provides distinct metadata to the information source to allow the sender and the recipient to verify that the information is accurate and cannot be modified or altered. The system also uses RNN-based LSTM and GRU model neural networks to determine whether the method performs better for these applications and uses an Ethereum blockchain to perform transactions between the two parties. The system also used an advanced encryption standard (AES) to protect the transactions. The AES uses encryption and decryption keys. This system structure is implemented on raspberry pi devices to determine its feasibility in a small system flowchart. The dataset for this study is the “Image encryption and decryption dataset” from the Kaggle website, and the results are better than GRU, with a 0.9635 accuracy rate for this dataset. The smart contract module is verifiable, available, tamper-resistant, and usable, and AES provides a more robust encryption method for the system.