Surrogate data testing is an important statistical method to detect nonlinearity in time series and has been widely used.The choice of test statistics can bring important influence to the nonlinearity of time series.A method for testing nonlinearity is described based on the information theory－redundancy.We use the AAFT algorithm to produce surrogate data and using redundancy as test statistics.The evaluation of redundancy statistics for univariate and multivariate is  described respectively.The result of numerical experiments confirms our approach is an effective and robust nonlinearity detecting method.