**The Promptware Kill Chain** logo

**The Promptware Kill Chain**

Free

FreeFree tier
Type
Open Source

About **The Promptware Kill Chain**

The Promptware Kill Chain is a research paper that reframes prompt injection attacks on large language models (LLMs) as a multistep malware delivery mechanism called 'promptware.' The authors propose a seven-stage kill chain—Initial Access, Privilege Escalation, Reconnaissance, Persistence, Command and Control, Lateral Movement, and Actions on Objective—based on analysis of 36 studies and real-world incidents. They demonstrate that at least 21 documented attacks traverse four or more stages, highlighting the need for a defense-in-depth approach across the entire life cycle. The paper provides a structured framework for risk assessment and security engineering of LLM-based systems.

Key Features

Seven-stage kill chain model: Initial Access, Privilege Escalation, Reconnaissance, Persistence, Command and Control, Lateral Movement, Actions on Objective
Analysis of 36 prominent studies and real-world incidents affecting production LLM systems
Identifies 21 documented attacks traversing four or more kill chain stages
Proposes defense-in-depth countermeasures for each stage
Shifts the conversation from isolated prompt injection to systematic promptware malware

Pros & Cons

Pros
  • Provides a clear, structured taxonomy for understanding advanced prompt injection attacks
  • Based on real-world incidents and published studies, not just theoretical scenarios
  • Offers specific countermeasures for each stage of the kill chain
  • Useful for both red-team penetration testing and blue-team defensive planning
Cons
  • Primarily a research framework, not a ready-to-use software tool
  • Requires manual adaptation and implementation by security teams
  • Some stages (e.g., lateral movement) may be less documented in practice
  • Does not provide automated detection or mitigation code

Best For

Structured risk assessment for LLM-based applicationsSecurity engineering and threat modeling of AI systemsAcademic research on LLM vulnerabilities and malware evolutionDesigning defense strategies against multistep prompt injection attacks

FAQ

What is promptware?
Promptware is a new class of malware execution mechanism that uses prompts engineered to exploit an application's LLM, evolving from simple prompt injection into a multistep attack chain.
How many stages are in the Promptware Kill Chain?
Seven: Initial Access (prompt injection), Privilege Escalation (jailbreaking), Reconnaissance, Persistence (memory and retrieval poisoning), Command and Control, Lateral Movement, and Actions on Objective.
What kind of countermeasures does the paper discuss?
It reviews countermeasures for each stage, advocating a defense-in-depth approach that addresses the entire promptware life cycle, though specific techniques are detailed only at a high level.