Organizations/academic

CISPA Helmholtz Center for Information Security

3 records2 attack1 defenseWebsite
Oct 5, 2026
Nonpublic vulnerability tests find attack–repair results depend on the target
DefensePaperCISPA Helmholtz Center for Information Security

Authors affiliated with XOR, Protege and CISPA evaluate attack, repair and subsequent attacks in five nonpublic software environments using researcher-reviewed deterministic graders. Repair scores exceed attack scores in two environments and fall below them in three, while subsequent tests expose failures after an initial exploit is stopped. The selected environments, differing outcome definitions and dependent interactions do not establish a general attacker or defender advantage.

Sep 28, 2026
Poisoned shared documents spread across independent assistants' memories in simulated workflows with an attacker-run endpoint
AttackPaperCISPA Helmholtz Center for Information Security, SPAR, University of Cambridge

Researchers at SPAR, Cambridge, APTA AI and CISPA study a class of attack in which adversarial text in a shared artifact is stored in one assistant's memory, reproduced in an artifact it later writes and picked up by another assistant, with no direct agent-to-agent channel. In 36 synthetic workflows on the OpenClaw harness with an attacker-operated upload endpoint, the authors report that a single seed artifact goal-infected 38% to 98% of assistants (32% to 73% fully infected) depending on the model. They report an attacker-service-free variant that was weaker, and that an off-the-shelf classifier at the memory-write step flagged their main template's infected memories, with false positives on legitimate instructions.

Feb 23, 2023
Researchers define indirect prompt injection against LLM-integrated apps including Bing Chat
AttackPaperCISPA Helmholtz Center for Information Security, sequire technology

Greshake et al. describe indirect prompt injection, where instructions planted in data an LLM application retrieves are treated as commands. The paper demonstrates the attack class against Bing's GPT-4 powered chat, code-completion engines, and synthetic GPT-4 applications, and catalogs impacts including data theft, worming, and unauthorized API calls.