72-Hour Phone Tracking Audit Challenges the Microphone Eavesdropping Theory
第404号工位 reports a 72-hour packet-capture audit on an iPhone 15 Pro and a rooted Google Pixel 8, claiming 41.8GB of traffic and no microphone hardware wake-ups or uploaded raw audio. The article attributes apparent prediction to spatial-temporal graph modeling, private set intersection and on-device embeddings rather than background recording. These findings and the broader claims about commercial tracking practices were not independently corroborated in the supplied material.
The report offers an AI-linked explanation for personalized recommendations that does not rely on microphone eavesdropping, but its unusual test setup and sweeping conclusions currently rest on one self-published account.
📋 实验审计档案 (Lab Audit Dossier #2026-NET05)
你一定经历过至少一次这种让人后背发凉的“灵异时刻”:
中午跟同事在工位吃便当,随口聊了一句:“最近后背酸得不行,想买个筋膜枪揉揉。” 在此之前,你没有在任何电商软件里搜过这三个字,没有在微信里打过这几个字,甚至连手机屏幕都没亮过。
然而当天下午下班,你在地铁上随手打开某个资讯或购物 App,首页信息流的第一条推荐,赫然就是一把黑色的筋膜枪。
那一瞬间,几乎所有人的第一反应都是: “绝对是手机麦克风在后台偷听我说话!”
#iPhone 15 Pro#Google Pixel 8#iOS 17.5#Android 14#mitmproxy#eBPF#Private Set Intersection#Spatial-Temporal GNN#On-Device Embedding