Regularization methods for high-dimensional variable selection and estimation have been intensively studied in recent years and most of them are developed in the framework of linear regression models where the predictor variables are assumed to be accurately measured. However, in real data analysis it is common that some predictors cannot be measured directly or precisely. While it is well known that measurement error in predictors causes attenuation in parameter estimation, its impact on variable section is not well studied. We study this problem in the framework of generalized linear models and propose an instrumental variable approach to correct the bias in variable section and estimation. We present some theoretical results as well as numerical examples.