Investigating the risk of linkability and ultimately identifiability from collected user data such as 1) raw EMG data, 2) filtered EMG data, and 3) matrices of data-driven models (model personalization parameters). We hope to show that filtering data increases the privacy protections, to identify which algorithms or classes of algorithms provide greater privacy protections, and to develop a framework for determining where the best point on the personalization-privacy may be for different applications, datasets, and algorithms related to human-computer interfaces.