Analyzing Defensive Misdirection Against Model-Guided Automated Attacks on Agentic AI Systems

{"Research":[0],"artifact":[1,18],"for":[2,36,47],"the":[3,21],"paper":[4],"\\"Analyzing":[5],"Defensive":[6],"Misdirection":[7],"AgainstModel-Guided":[8],"Automated":[9],"Attacks":[10],"on":[11],"Agentic":[12],"AI":[13],"Systems\\"":[14],"(ACSAC":[15],"2026).":[16],"This":[17],"contains:":[19],"(1)":[20],"LLMaaD":[22],"defense":[23],"framework":[24],"implementing":[25],"theCMPE":[26],"(Confuse,":[27],"Misdirect,":[28],"Preserve,":[29],"Evade)":[30],"algorithm;":[31],"(2)":[32],"integration":[33],"andevaluation":[34],"scripts":[35],"AutoDAN-Turbo,":[37],"AutoDAN-Reasoning,":[38],"GPTFuzz,":[39],"and":[40],"PAIRattack":[41],"frameworks;":[42],"(3)":[43],"pre-computed":[44],"result":[45],"CSVs":[46],"all":[48],"experiments.":[49]}