Decrypting Chaotic Visual Ciphers via Hybrid Quasi Quantum Neural Networks (Q²NNs)

This repository provides a fully reproducible experimental pipeline for reconstructing chaotically transformed MNIST images using a hybrid Quantum–Classical Neural Network (Q²NN). The workflow integrates a classical CNN autoencoder, a variational QNN, and a fusion model combining both branches. Encrypted datasets, modular Jupyter notebooks, and all code are provided to enable step-by-step replication of the experiments, including training, evaluation, and visualization of reconstruction performance. Deterministic seeds and fixed encryption keys ensure reproducibility; minor numerical differences may occur due to stochastic initialization and simulator variability.