Tick Stream Tampering-Resistant Anomaly Detection in Stock Market using Integrated Time Series and Behavioral Analysis

This study explores cutting-edge methods for detecting anomalies in the stock market by combining deep learning techniques with insights from behavioral finance. Using models like LSTM-based autoencoders and Graph Neural Networks (GNNs), we aim to uncover unusual patterns in stock prices and trading behavior. The research also simulates various scenarios, such as Tickstream attacks where real-time trading data is manipulated and market spoofing where massive buy/sell orders distort prices, to test the system's ability to detect fraud. By incorporating behavioral factors, like investor sentiment, we provide a deeper understanding of how emotions influence the market. The results show that our approach improves market monitoring, helps prevent fraud, and gives valuable insights into both technical anomalies and behavior-driven market shifts. This framework can be applied in real-time, making it a powerful tool for financial surveillance.