Seeing is Deceiving: Using Adversarial Images to Prevent VLM-Based Plagiarism