Generating Captcha Images Using Variations of the Same Object

Aspects described herein may allow for the application of generating captcha images using variations of the same object. A GAN model may generate objects and backgrounds of the captcha images and the model may be trained based on ground-truth images to obtain refined model parameters. Further aspects described herein may provide for generating variants of the objects based on the trained model and the refined model parameters. The synthetic captcha images may be created based on the backgrounds and variants of the objects. Finally, the synthetic captcha images and ground-truth images may be presented as security challenges and user access requests may be granted or denied based on responses to the security challenges.