The Devil is in the Context: Nine Dimensions of Prompt Injection Attacks

Evidence record

Read the original source ↗

Claim-level citations (3)

• supportsPrompt-injection benchmarks missed attack combinations despite large sample counts: The study found 19 of 144 pairwise attack-property combinations absent across nine benchmarks, including 15 of 36 rare-property pairings.

  Section V-B Coverage Analysis and nine-dimensional pairwise table

• supportsPrompt-injection benchmarks missed attack combinations despite large sample counts: On 40 generated AgentDojo variants, CaMeL recorded 2.5% attacker success and 27.5% legitimate utility, while PromptArmor recorded 7.5% attacker success and 90% legitimate utility.

  Section VI, Security and utility results, Figure 6

• supportsPrompt-injection benchmarks missed attack combinations despite large sample counts: On 40 selected AgentDojo Workspace instances, generated variants achieved 17.5% attack success without defense while the selected original cases recorded zero for the reported test configuration.

  Section VI-B Taxonomy Guided Case Generation; Security and utility results