BackgroundCybervictimization and cyberaggression are increasingly recognized as public health concerns among university students; however, evidence from Thailand remains limited.ObjectiveThis study examined whether cybervictimization was associated with cyberaggression endorsement and whether self-esteem moderated this association. We used a two-part zero-inflated Poisson (ZIP) framework to distinguish between (a) a structural-zero/no-endorsement process and (b) the extent of cyberaggression endorsement among individuals at risk.MethodsA cross-sectional online survey was conducted in January 2023 among university students in Chiang Mai, Thailand (N = 414). Cybervictimization was assessed using the victimization subscale of the Thai version of the Cyberaggression–Perpetration and Victimization Scale. Cyberaggression was operationalized as a bounded behavioral endorsement count (AGGRESSO; range = 0–5), reflecting the number of cyberaggressive perpetration behaviors endorsed during the past month. Self-esteem was assessed using the Thai Rosenberg Self-Esteem Scale. Competing Poisson, negative binomial, and ZIP models were compared using information criteria.ResultsModel comparisons supported the ZIP specification (AIC = 484.143; BIC = 516.350). In the ZIP count component, higher self-esteem was associated with a lower extent of cyberaggression endorsement among at-risk individuals (b = −0.078, p = 0.005). Cybervictimization (b = 0.048, p = 0.386) and the cybervictimization × self-esteem interaction (b = 0.015, p = 0.180) were not statistically significant in this component.In the zero-inflation component, the cybervictimization × self-esteem interaction was statistically significant (b = −0.106, p = 0.050), while the main effect of cybervictimization was marginal (b = −1.406, p = 0.054). These findings indicate that the association between cybervictimization and structural non-endorsement—equivalently, the likelihood of any cyberaggression endorsement—differed according to self-esteem. Sensitivity analyses using binary and ordinal outcome specifications produced a consistent pattern: cybervictimization and its interaction with self-esteem were most evident for the involvement aspect of cyberaggression (i.e., any endorsement), rather than the extent of endorsement.ConclusionCybervictimization was more strongly related to cyberaggression involvement than to the extent of cyberaggression endorsement among at-risk individuals, whereas self-esteem showed a component-specific pattern. These findings demonstrate the value of distinguishing between involvement and endorsement extent in cyberaggression research. University prevention efforts may benefit from timely support following cybervictimization, together with interventions that strengthen self-regulation and adaptive coping resources.