AI-Enhanced Rear-End Collision Detection Using Attention-LSTM with Predictive and Cybersecurity Capabilities

Rear-end collisions are one of the most common types of traffic accidents, often caused by insufficient reaction times or driver inattention. In recent years, integrating artificial intelligence (AI) in the automotive industry has revolutionized safety measures. In this paper, we propose an AI-driven rear-end collision detection system using an Attention-based Long Short-Term Memory (Attention-LSTM) network, designed to enhance predictive capabilities and integrate robust cybersecurity features. The model leverages vehicle sensor data, including speed, acceleration, brake pressure, and distance to the preceding vehicle, to predict potential rear-end collisions. Including an attention mechanism allows the model to focus on critical time steps that are most indicative of an impending collision, significantly improving both short-term and long-term predictive accuracy. In addition to predictive performance, the system is fortified with a hybrid encryption model that secures data transmission between vehicles, ensuring resilience against common cyber threats such as spoofing and data tampering. This dual-layer approach addresses both the safety and cybersecurity needs of modern intelligent transportation systems. Experimental results demonstrate that the proposed Attention-LSTM model outperforms traditional models, including standard LSTM and Support Vector Machines (SVM), in terms of accuracy and responsiveness. The system's cybersecurity features also prove effective in defending against potential attacks without compromising real-time performance.