多模态钓鱼网站检测系统，支持任意客户端调用不同模态的专家模型。
Role-Aware Multi-modal federated learning system for detecting phishing webpages
- 采用硬门控专家机制，按输入模态选择对应专家进行参数独立聚合。
- 在TR-OP数据集上达到97.5%准确率，误报率仅2.4%。
- 适合隐私敏感场景下的跨设备、灵活可扩展的网页安全检测。
我们提出一种联邦多模态钓鱼网站检测系统，支持URL、HTML和图像输入，且客户端在推理时无需绑定固定模态，可调用其他客户端训练的任意模态头。方法上，在FedProx基础上提出角色感知桶聚合，受Mixture-of-Experts与FedMM启发，摒弃可学习路由，改用硬门控（根据样本模态选择图像/文本专家），实现模态专属参数的分离聚合，有效避免跨嵌入冲突并稳定收敛。在TR-OP数据集上，融合头实现97.5%准确率，误报率2.4%，覆盖两种数据类型；在图像子集上达95.5%准确率，误报率5.9%。文本方面，对URL使用GraphCodeBERT，对原始噪声HTML采用早期三路嵌入。在WebPhish（HTML）上获得96.5%准确率、1.8%误报率；在TR-OP（原始HTML）上达95.1%准确率、4.6%误报率。结果表明，硬门控桶聚合在严格隐私保护下实现了稳定联邦训练，并提升了多模态钓鱼检测的可用性与灵活性。
原文摘要 · Abstract (English)
We present a federated, multi-modal phishing website detector that supports URL, HTML, and IMAGE inputs without binding clients to a fixed modality at inference: any client can invoke any modality head trained elsewhere. Methodologically, we propose role-aware bucket aggregation on top of FedProx, inspired by Mixture-of-Experts and FedMM. We drop learnable routing and use hard gating (selecting the IMAGE/HTML/URL expert by sample modality), enabling separate aggregation of modality-specific parameters to isolate cross-embedding conflicts and stabilize convergence. On TR-OP, the Fusion head reaches Acc 97.5% with FPR 2.4% across two data types; on the image subset (ablation) it attains Acc 95.5% with FPR 5.9%. For text, we use GraphCodeBERT for URLs and an early three-way embedding for raw, noisy HTML. On WebPhish (HTML) we obtain Acc 96.5% / FPR 1.8%; on TR-OP (raw HTML) we obtain Acc 95.1% / FPR 4.6%. Results indicate that bucket aggregation with hard-gated experts enables stable federated training under strict privacy, while improving the usability and flexibility of multi-modal phishing detection.
Thank you to arXiv for use of its open access interoperability. PaperDance 不是 arXiv 官方产品；中文卡片由大模型生成，请以原文为准。