We present a trustworthiness score for double-sided two-way ranging on ultra-wideband devices. It serves as quantitative measure for assessing the reliability and security aspects of ranging. Leveraging the trustworthiness score, we propose two localization schemes that utilize the weighted non-linear least squares approach. The computation of the trustworthiness score relies on an autoencoder model trained on trustworthy data. To evaluate the performance of our trustworthiness score, we conduct indoor localization experiments that include channel manipulations such as line-of-sight obstructions and timestamp attacks resembling the well-known Cicada attack. The incorporation of trustworthiness led to an improvement in localization accuracy, reducing the root mean square error (RMSE) by up to 50% in the dynamic scenario.