Previous studies suggested that social anxiety is associated with interpretation bias, theory of mind deficit, and eye gaze avoidance when identifying facial emotions. We tested the hypothesis that socially anxious individuals would be more affected by mask use during facial emotion recognition. 88 healthy undergraduates with various levels of social anxiety were invited. Participants judged the emotions of masked and unmasked facial expressions. Eye Movement Analysis with Hidden Markov Models was used to analyze participants' eye movement patterns during the task. Potential confounders including gender, depressive symptoms, stress, and executive planning ability were controlled for in the analyses. Results failed to support our hypothesis. Instead, higher social anxiety was associated with higher accuracy rates for angry and fearful faces and lower false alarm rates for sad faces. Eye movement patterns were similar across social anxiety levels. Interestingly, an exploratory moderation analysis revealed that an increase in using a more eye-centered strategy due to mask use was significantly associated with a larger drop in accuracy rate for fearful faces among individuals with higher social anxiety, while non-significantly associated with a smaller drop among individuals with lower social anxiety. Thus, our study indicates social anxiety, at least at subclinical levels, may be associated with a generally heightened sensitivity to negative emotions. However, such heightened sensitivity diminishes if they switch to a more eye-centered strategy when viewing masked facial emotions. Potential mechanisms and implications were discussed.