Visual small target motion detection is crucial for time-critical applications such as real-time UAV surveillance, autonomous robotics perception, and target tracking. However, conventional frame-based cameras are fundamentally limited by low sampling rates and motion blur. Event cameras, which provide asynchronous outputs with microsecond temporal resolution by recording only brightness changes, offer a promising alternative. Despite these advantages, event-based small target detection remains challenging due to the incompatibility of image-processing paradigms and the dominance of noise events that obscure sparse small target motion signals. Inspired by the spike-based processing of insect small target motion detector (STMD) neurons, which naturally align with event data, we propose a novel STMD-based spiking neural network model (sSTMD) to address the incompatibility issues. To further enhance noise suppression, we introduce a Minkowski-chain dendritic summation mechanism integrated into the sSTMD, which leverages strict spatiotemporal causality—modeled via Minkowski spacetime intervals—to distinguish coherent target motion from uncorrelated noise. Theoretical analysis and numerical experiments on a dedicated dataset demonstrate that our method robustly detects small target motion while significantly outperforming existing models in noise suppression. This work not only provides an effective bio-inspired solution for event-based small target detection but also introduces a principled, causality-driven approach to enhancing noise resilience in spiking neural networks.