Can Watermarked LLMs be Identified by Users via Crafted Prompts?