Breaking Down the Defenses: A Comparative Survey of Attacks on Large Language Models