QGuard:Question-Based Zero-Shot Guard for Multi-Modal LLM Safety

The recent advancements in Large Language ModelsLLMs have had a significant impact on a wide range of fields, from general domains to specialized areas. However, these advancements have also significantly increased the potential for malicious users to exploit harmful and jailbreak prompts for...

NAP-Tuning: Neural Augmented Prompt Tuning for Adversarially Robust Vision-Language Models

Vision-Language Models VLMs such as CLIP have demonstrated remarkable capabilities in understanding relationships between visual and textual data through joint embedding spaces. Despite their effectiveness, these models remain vulnerable to adversarial attacks, particularly in the image modality,...

Multi-Modal Multi-Task Federated Foundation Models for Next-Generation Extended Reality Systems: Towards Privacy-Preserving Distributed Intelligence in AR/VR/MR

Extended reality XR systems, which consist of virtual reality VR, augmented reality AR, and mixed reality XR, offer a transformative interface for immersive, multi-modal, and embodied human-computer interaction. In this paper, we envision that multi-modal multi-task M3T federated foundation model...

BadReward: Clean-Label Poisoning of Reward Models in Text-To-Image RLHF

Reinforcement Learning from Human Feedback RLHF is crucial for aligning text-to-image T2I models with human preferences. However, RLHF's feedback mechanism also opens new pathways for adversaries. This paper demonstrates the feasibility of hijacking T2I models by poisoning a small fraction of...

ECG Identity Authentication in Open-Set with Multi-Model Pretraining and Self-Constraint Center and Irrelevant Sample Repulsion Learning

Electrocardiogram ECG signal exhibits inherent uniqueness, making it a promising biometric modality for identity authentication. As a result, ECG authentication has gained increasing attention in recent years. However, most existing methods focus primarily on improving authentication accuracy...

Indirect Instruction Injection in Multi-Modal LLMs

Interesting research: "Abusing Images and Sounds for Indirect Instruction Injection in Multi-Modal LLMs": Abstract: We demonstrate how images and sounds can be used for indirect prompt and instruction injection in multi-modal LLMs. An attacker generates an adversarial perturbation corresponding t...

Corporate Office and Kitchen Table: Securing the Future of Work, Part 1

The future of work is multi-modal, the future corporate office is a private coffee shop with great Wi-Fi, and the future of enterprise security is going to have to adapt rapidly. If there is a sliver of positivity that I can find in this devastating pandemic, it's that we are adapting and finding...

Researchers Uncover Novel Way to De-anonymize Device IDs to Users' Biometrics

Researchers have uncovered a potential means to profile and track online users using a novel approach that combines device identifiers with their biometric information. The details come from a newly published research titled "Nowhere to Hide: Cross-modal Identity Leakage between Biometrics and...

A week in security (May 28 – June 3)

Last week on Labs, we talked about the significance of SEO poisoning in the world of search marketing, blackmail attempts against financial institutions in Canada, voice command flaws in smart assistants, survey and potential phishing scams on Instagram, and the latest changes in Office 365. We...