Deep Learning-Based Financial Forecasting with Post-Quantum Cryptographic Integration: A CNN-LSTM and F2N-ECC Hybrid Framework

This study presents a secure-by-design deep learning framework that combines CNN-LSTM-based forecasting with post-quantum cryptographic protection using Field-to-Number Elliptic Curve Cryptography (F2N-ECC). In contrast to earlier research that often relied on legacy financial data, our evaluation is conducted on a modern, high-frequency dataset that includes hourly cryptocurrency prices, global equity indices, and sentiment-based market indicators. The proposed architecture enables encrypted forecasting to be performed directly on sensitive financial inputs, preserving data confidentiality throughout the pipeline. Despite operating on encrypted data, the model achieves strong predictive performance, with an R² score of approximately 0.89. Experimental results also highlight significant efficiency gains - F2N-ECC reduces encryption latency by up to 50% compared to traditional ECC methods. Furthermore, the system demonstrates robustness under highly volatile market conditions, validating its suitability for real-world, high-risk financial forecasting tasks.