Knowledge Base Regulation Tracker Standards & Frameworks OSS Directory AI Incident Database Research Paper Library Case Studies Society Articles News Jobs About Charter Sign inBecome a Member
Home/Research Paper Library/Abusing Images and Sounds for Indirect Instruction Injection in Multi-Modal LLMs

Abusing Images and Sounds for Indirect Instruction Injection in Multi-Modal LLMs

Extended indirect prompt injection to images and audio, showing multimodal models can be hijacked by perturbed media.

multimodalimage injectionaudio injection

Contribute

Spot an error, or know a record we should add? Members can propose additions and submit corrections.

Submit researchReport a correction or request removal

Related in Research Paper Library

TERMon: Detecting Persistent Behavioral Threats in Edge AI via Hardware-Native Ternary Runtime Monitor

Edge AI accelerators are increasingly deployed in safety-critical environments, where model outputs may control physical actuators, make access-control decisions, or trigger alarms. In these settings, runtime…

Multimodal and physical-world attacks2026

Fingerprinting Multimodal Large Language Models

While multimodal large language models (MLLMs) enable a wide range of image-text reasoning tasks, recent incidents indicate that they are vulnerable to illicit deployment and unauthorized distillation. Existing…

Multimodal and physical-world attacks2026

Weather Data Spoofing Attacks on Rain-Adaptive Millimeter-Wave Frequency Selection in V2X Communication Networks

Connected vehicles use millimeter-wave (mmWave) sidelinks for the data rates cooperative driving demands, and emerging designs select the carrier band from sensed rainfall. We show that this weather awareness is an…

Multimodal and physical-world attacks2026