Today’s widespread use of stroke-based touchscreen devices creates numerous associated security concerns and requires efficient security measures in response. We propose an imitation-resistant passive authentication interface for stroke-based touch screen devices employing classifiers for each individual stroke, which is evaluated with respect to 26 features. For experimental validation, we collect stroke-based touchscreen data from 23 participants containing target and imitation stroke patterns using a photo-matching game in the form of an iOS application. The equal error rate (EER), depicting the rate at which false rejection and false acceptance of target and imitator strokes are equal, is assumed as an indicator of the classification accuracy. Leave-one-out cross-validation was employed to evaluate the datasets based on the mean EER. For each cross-validation, one out of the two target datasets, an imitator dataset, and the remaining 20 imitator datasets were selected as genuine data, imitator test data, and imitator training data, respectively. Our results confirm stroke imitation as a serious threat. Among the 26 stroke features evaluated in terms of their imitation tolerance, the stroke velocity was identified as the most difficult to imitate. Dividing classifiers based on the stroke direction was found to further contribute to classification accuracy.