Adversarial Attacks on Large Language Model-Based System and Mitigating Strategies: A Case Study on ChatGPT | Synapse