Faster-GCG: Efficient Discrete Optimization Jailbreak Attacks against Aligned Large Language Models