Cyber-physical systems (CPSs) are complex interconnections of control systems that operate in tandem to carry out industrial processes and activities. A research problem of interest is the lack of cyber-attack models for a CPS, which hinders the design of robust defense mechanisms. We aim to generate variant attack patterns through the application of association rule mining to a commonly available dataset, namely SWaT. Our methodology comprises the generation of synthetic attack variants to foster the development of robust intrusion detection models. Synthetic samples created by this process can be used to create adversarial samples, which have not yet been addressed in the field of CPS. In this work, we have also generated more than six hundred classification rules and applied time series analysis to validate potential data injection attack variants for the SWaT dataset.