A method is presented for real-time validation of GNSS measurements of a single receiver, where data from each satellite are independently processed. A geometry-free observation model is used with a reparameterized form of the unknowns to overcome rank deficiency of the model. The ionosphere error and non-constant biases such as multipath are assumed changing relatively smoothly as a function of time. Data validation and detection of errors is based on statistical testing of the observation residuals using the Detection-Identification-Adaptation (DIA) approach. The method is applicable to any GNSS with any number of frequencies. The performance of validation method was evaluated using multiple-frequency data from three GNSS (GPS, GLONASS and Galileo) that span three days in a test site at Curtin University, Australia. Performance of the method in detection and identification of outliers in code observations and detection of cycle slips in phase data was examined. Results show that the success rate vary according to precision of observations and their number as well as size of the errors. The method capability is demonstrated when processing four IOV Galileo satellites in a single point positioning mode, and in another test by comparing its performance with Bernese software in detection of cycle slips in PPP processing using GPS data.