We introduce an incremental total least-squares vehicle mass estimation algorithm, based on a vehicle longitudinal dynamics model. Available control area network signals are used as model inputs and output. In contrast to common vehicle mass estimation schemes, where noise is only considered at the model output, our algorithm uses an errors-in-variables formulation and considers noise at the model inputs as well. A robust outlier treatment is realized as batch total least-squares routine and hence, the proposed algorithm works in a superior way on a broad range of vehicle acceleration. The results of six test runs on various vehicle masses show highly accurate mass estimation results on high and low dynamics of vehicular operation.