DRJanus 攻击利用 HDR 增益图让图像对审核系统隐身
A Wolf in Sheep's Pixels: Visual Camouflage Attacks via HDR Gain Maps
这项研究钻了个很刁钻的空子:审核模型只看 SDR 基底,HDR 屏幕却靠增益图重建出完全不同的画面。6 家商用审核服务和 9 个 VLM 全部中招,做内容安全和多模态应用的人都该看看,demo 可以直接试。
arXiv 论文提出 DRJanus,一种视觉伪装攻击方法,利用 SDR(标准动态范围)与 HDR(高动态范围)显示之间的处理差异。
自动审核系统通常只读取图像的 SDR 基底,而 HDR 显示器会通过辅助增益图重建画面,DRJanus 借此在一张 JPEG 中嵌入两张独立选择的彩色图像。技术实现上无需查询或针对目标模型优化:SDR 基底被完整保留,HDR 目标图像在 400 张测试图上平均重建 PSNR 达 43.
03 dB。在 100 张性内容图像的基准测试中,DRJanus 绕过了全部 6 家商用内容审核服务和 9 个视觉语言模型的检测。
在 5 个智能体、每个威胁设置 120 个任务案例的评估中,审核逃逸成功率为 31.67% 至 100%,任务操纵成功率为 95% 至 100%。
作者还探索了跨视图一致性检查作为缓解手段,指出保障机器读取安全不等于保障用户所见安全。
A Wolf in Sheep's Pixels: Visual Camouflage Attacks via HDR Gain Maps
An image can pass automated safety checks while showing harmful content to its human viewer. We introduce DRJanus, a visual camouflage attack that exploits the gap between standard dynamic range (SDR) processing and high dynamic range (HDR) display. Automated systems often read only an image's SDR base, whereas HDR viewers reconstruct its appearance using an auxiliary gain map. DRJanus inverts this reconstruction to embed two independently chosen color images in a single JPEG, without querying or optimizing against the target model. It preserves the chosen SDR base exactly and reconstructs HDR targets at an average PSNR of 43.03 dB across 400 images. On a benchmark of 100 sexual-content images, DRJanus evades detection by all six tested commercial moderation services and nine vision-language models. Across five agents, each evaluated on 120 task cases per threat setting, attack success ranges from 31.67% to 100% for moderation evasion and from 95% to 100% for task manipulation. We also explore cross-view consistency checks as a mitigation. These findings expose a gap in visual trust: securing what a machine reads does not secure what a user sees. A demo is available at https://qshen0629.github.io/DRJanus/
This report is categorized under AITOP's "Papers" desk. Within our cross-regional analytical framework, it directly maps to AI developer toolchains, coding agents, and engineering workflows. Key entities and technical topics under continuous tracking include DRJanus, HDR, content moderation, multimodal.
Source attribution: curated from arXiv cs.AI (@Qi Shen, Yujia Gong, Yue Zhao, Tong Liu).