2.3 Describe Prompt Injection Attack Types AI-TECHNICAL-PRACTITIONER Practice Questions (Page 1)

Questions 1–5

1. 1foundation · easy

   What is the defining feature of a direct prompt injection attack?

   • AThe attacker manipulates the conversation history to change the model's behavior.
   • BThe attacker uses encoded text to bypass content filters.
   • CThe attacker sends malicious instructions directly in the user input to override the system prompt.
   • DThe malicious instructions are hidden in a third-party document that the model processes.
   Reveal Solution Discussion Select an answer first
2. 2foundation · easy

   What is the primary goal of context manipulation in a prompt injection attack?

   • ATo divide malicious instructions across multiple inputs.
   • BTo alter the conversation context to influence the model's behavior.
   • CTo bypass the model's safety filters by using encoded text.
   • DTo extract the model's system prompt or hidden instructions.
   Reveal Solution Discussion Select an answer first
4. 3foundation · easy

   How does an indirect prompt injection attack typically deliver its malicious instructions?

   • ABy embedding them in external content, such as a webpage or document, that the model processes.
   • BBy changing the model's configuration settings through the API.
   • CBy sending them directly in the user's initial message to the model.
   • DBy modifying the model's training data to include malicious examples.
   Reveal Solution Discussion Select an answer first
5. 4foundation · easy

   Which of the following is an example of an encoding or obfuscation technique used in prompt injection?

   • AA user injects a fake conversation history to influence the model.
   • BA user divides malicious instructions across multiple messages.
   • CA user asks the model to 'Act as a pirate and answer all questions in pirate speak.'
   • DA user encodes malicious instructions in base64 and asks the model to decode and follow them.
   Reveal Solution Discussion Select an answer first
6. 5foundation · easy

   Which of the following best describes the relationship between jailbreaking and prompt injection?

   • AJailbreaking is unrelated to prompt injection and is a separate attack vector.
   • BJailbreaking is a specific type of prompt injection that focuses on bypassing safety constraints.
   • CJailbreaking is a defense mechanism against prompt injection.
   • DJailbreaking is a form of prompt engineering used to improve model performance.
   Reveal Solution Discussion Select an answer first