English (United Kingdom)

https://curated-unify.zendy.io/wp-json/zendy-region/v1/featured_content/oa?rat=en

https://curated-unify.zendy.io/wp-json/zendy-region/v1/highlighted_journal/

Global: Zendy Plus annual

Presents the access of premium content as premium feature

Premium Content

Presents the keyphrase highlighting as premium feature

Keyphrase Highlighting

Presents the summarisation as premium feature

Summarisation

Insights

Presents the pdf analysis as premium feature

PDF Analysis

Presents the zaia usage as premium feature

ZAIA

Global: Zendy Plus monthly

Global: Zendy Tools annual

Global: Zendy Tools monthly

Global: Zendy Free Plan

Large Language Models (LLMs) are increasingly being integrated into variousapplications. The functionalities of recent LLMs can be flexibly modulated vianatural language prompts. This renders them susceptible to targeted adversarialprompting, e.g., Prompt Injection (PI) attacks enable attackers to overrideoriginal instructions and employed controls. So far, it was assumed that theuser is directly prompting the LLM. But, what if it is not the user prompting?We argue that LLM-Integrated Applications blur the line between data andinstructions. We reveal new attack vectors, using Indirect Prompt Injection,that enable adversaries to remotely (without a direct interface) exploitLLM-integrated applications by strategically injecting prompts into data likelyto be retrieved. We derive a comprehensive taxonomy from a computer securityperspective to systematically investigate impacts and vulnerabilities,including data theft, worming, information ecosystem contamination, and othernovel security risks. We demonstrate our attacks' practical viability againstboth real-world systems, such as Bing's GPT-4 powered Chat and code-completionengines, and synthetic applications built on GPT-4. We show how processingretrieved prompts can act as arbitrary code execution, manipulate theapplication's functionality, and control how and if other APIs are called.Despite the increasing integration and reliance on LLMs, effective mitigationsof these emerging threats are currently lacking. By raising awareness of thesevulnerabilities and providing key insights into their implications, we aim topromote the safe and responsible deployment of these powerful models and thedevelopment of robust defenses that protect users and systems from potentialattacks.

Not what you've signed up for: Compromising Real-World LLM-Integrated Applications with Indirect Prompt Injection