Bridging the Gap in Vision Language Models in Identifying Unsafe Concepts Across Modalities
July 15, 2025 Β· Declared Dead Β· π USENIX Security Symposium
"No code URL or promise found in abstract"
Evidence collected by the PWNC Scanner
Authors
Yiting Qu, Michael Backes, Yang Zhang
arXiv ID
2507.11155
Category
cs.CR: Cryptography & Security
Cross-listed
cs.AI
Citations
1
Venue
USENIX Security Symposium
Last Checked
4 months ago
Abstract
Vision-language models (VLMs) are increasingly applied to identify unsafe or inappropriate images due to their internal ethical standards and powerful reasoning abilities. However, it is still unclear whether they can recognize various unsafe concepts when presented in different modalities, such as text and images. To address this, we first compile the UnsafeConcepts dataset, featuring 75 unsafe concepts, i.e., ``Swastika,'' ``Sexual Harassment,'' and ``Assaults,'' along with associated 1.5K images. We then conduct a systematic evaluation of VLMs' perception (concept recognition) and alignment (ethical reasoning) capabilities. We assess eight popular VLMs and find that, although most VLMs accurately perceive unsafe concepts, they sometimes mistakenly classify these concepts as safe. We also identify a consistent modality gap among open-source VLMs in distinguishing between visual and textual unsafe concepts. To bridge this gap, we introduce a simplified reinforcement learning (RL)-based approach using proximal policy optimization (PPO) to strengthen the ability to identify unsafe concepts from images. Our approach uses reward scores based directly on VLM responses, bypassing the need for collecting human-annotated preference data to train a new reward model. Experimental results show that our approach effectively enhances VLM alignment on images while preserving general capabilities. It outperforms baselines such as supervised fine-tuning (SFT) and direct preference optimization (DPO). We hope our dataset, evaluation findings, and proposed alignment solution contribute to the community's efforts in advancing safe VLMs.
Community Contributions
Found the code? Know the venue? Think something is wrong? Let us know!
π Similar Papers
In the same crypt β Cryptography & Security
R.I.P.
π»
Ghosted
R.I.P.
π»
Ghosted
The Limitations of Deep Learning in Adversarial Settings
R.I.P.
π»
Ghosted
Distillation as a Defense to Adversarial Perturbations against Deep Neural Networks
R.I.P.
π»
Ghosted
Spectre Attacks: Exploiting Speculative Execution
R.I.P.
π»
Ghosted
How To Backdoor Federated Learning
R.I.P.
π»
Ghosted
Evasion Attacks against Machine Learning at Test Time
Died the same way β π» Ghosted
R.I.P.
π»
Ghosted
Federated Learning: Strategies for Improving Communication Efficiency
R.I.P.
π»
Ghosted
In-Datacenter Performance Analysis of a Tensor Processing Unit
R.I.P.
π»
Ghosted
Deep Convolutional Neural Networks for Computer-Aided Detection: CNN Architectures, Dataset Characteristics and Transfer Learning
R.I.P.
π»
Ghosted