扁平化手绘风技术人员科研人员研究人员工作场景图标元

生成一张TASLP论文Introduction用“Motivation chain”图，强调本文动机与独特价值，不做与他人方法对比，不出现“naive/benchmark”等字样。风格：极简矢量扁平、白底、少文字、可黑白打印。布局：横向1×4四格，从左到右标(1)(2)(3)(4)，每格标题1行+少量标签。(1) Title: Why ADD：画电话/客服/金融图标+红色“Deepfake”警示；标签：Need robust deepfake detector.(2) Title: Privacy Paradox：画大量“real speech waveform”文件+锁/盾牌；箭头指向云端被红叉阻断；再加小注：Centralizing speech violates privacy/compliance & may be misused for voice cloning.(3) Title: Our Key Idea：画多个客户端各自本地训练，raw audio留在本地（标data stays local）；云端只接收“model updates + class prototypes”；在云端中心写一句：Consensus learning under privacy constraint（强调从多客户端达成真实性共识、抑制偏置与不稳定）。不要画具体模块名。(4) Title: Outcome：画云端输出一个大图标Robust Deepfake Detector，旁边绿色对勾：High accuracy & stable；小字：No raw audio sharing. 底部整图一句话：Goal: High-performance ADD without sharing raw speech.