Support vector machine approach is an effective technique to solve poly-dimensional outlier detection, which can avoid the curse of dimensionality problem and has higher accuracy. One-class support vector machine-based outlier detection techniques take advantage of spatial and temporal correlations that exist between sensor data to cooperatively identify outliers. However, for large scale training samples, SVM techniques take more spatial and temporal overhead to process and optimize training samples. In this paper, we propose KNN-SVM techniques (Support Vector Machines based on K-Nearest Neighbor Algorithm) for Outlier Detection in Wireless Sensor Networks. It utilizes KNN techniques to reduce training samples' scale which can shorten training time and optimize time. Then it maps the samples into feature space by kernel function. Experiments with data collected from the Intel Berkeley Research Laboratory show that our techniques are feasible and can effectively reduce spatial and temporal consumption with high accuracy.