Adversarial-resilient iot-enabled financial intelligence system using explainable deep learning for sentiment-driven stock trading decisions

Abstract

Financial stock market trend decision systems support the investment decisions by predicting future price movements and analyzing large-scale market data. Despite, IoT enabled trading systems continuously collects and exchange sentiment and market information are exposed to adversarial attacks, which misleads automated trading algorithms and results in inaccurate investment decision systems. To face this challenge, this study proposes an adversarial-resilient explainable artificial intelligence for sentiment-driven stock trading (ARXAI-SST) architecture. The proposed framework combines graph neural networks (GNN) for adversarial attack detection, FinBERT based sentiment representation, BiLSTM with attention (BiLSTM-Att) for the sentiment analysis, and SHAP-based explainable artificial intelligence for enhancing model interpretability. The experimentation outcomes show that the proposed framework accomplishes higher accuracy and robustness compared with the existing approaches while providing transparent insights into the trading decisions. The proposed model offers an efficient and reliable solution for secure sentiment driven stock trading in IoT enabled financial environments.