
今年夏天,我的团队购买了三个月的商业可用数据,以了解某人——或许是某个心怀不轨的人——能从中获知美军官兵的哪些信息。我们从一个与我们一些最敏感的特种作战单位公开相关的军事设施入手。我们提取了在那里出现的广告 ID。共有数万个。随后我们寻找同时也出现在另外两个公开已知特种作战地点的 ID。只剩下几百个。
This summer, my team bought three months of commercially available data to see what someone — perhaps someone ill-intentioned — could learn about American troops. We started with a military installation publicly associated with some of our most sensitive special operations units. We pulled the advertising IDs seen there. There were tens of thousands. We then looked for IDs that also appeared at two other publicly known special operations locations. A few hundred remained.
我们挑选了一部夜间停留在一处住宅的手机。公开的财产记录和社交媒体将我们引向一名被公开描述为从事后勤工作的陆军军官。他的行动轨迹却透露了别的信息。我们的研究人员最终证实,他的实际职务是对手极其关心的那种。再点击几下,就看到了他孩子的照片。那是他的私人手机。我们没有黑进手机,没有入侵政府网络,也没有做任何违法之事。
We picked one phone that spent its nights at a house. Public property records and social media led us to an Army officer publicly described as working in logistics. His movements told us something else. Our researchers eventually confirmed that his actual role was one about which an adversary would care a great deal. A few more clicks led to photos of his children. The phone was his personal phone. We did not hack it. We did not compromise a government network or do anything illegal.
我们没有传唤无线运营商。我们只需购买商业可用数据。这本应让人人感到恐惧。这次实验让我感到不安,理应如此。我们正在注视一名美国现役军人以及间接地其家人的行踪。一个不受我们宪法约束的外国情报机构,可以购买我们购买的大部分相同数据。私营公司、犯罪组织或任何能进入该市场的人也能做到。当我担任“绿色贝雷帽”特种部队队员时,通过手机定位某人需要昂贵的设备和能力,这些在很大程度上属于政府。
We did not subpoena a wireless carrier. All we had to do was buy commercially available data. That should terrify everyone. The exercise made me uncomfortable, and rightly so. We were looking at the movements of an American service member and, indirectly, his family. A foreign intelligence service, not operating under our Constitution, can buy much of the same data we bought. So could a private company, a criminal organization or anyone else with access to the market. When I served as a Green Beret, finding someone through his phone required expensive equipment and capabilities that largely belonged to governments.
如今,一台笔记本电脑加上任何人都能购买的数据集,就能让你惊人地接近目标。AI 让问题变得更糟。它能筛选海量数据集,识别出过去需要分析师花费数天才能拼凑出的模式。斯坦福大学的 AI 指数显示,2022 年底至 2024 年底,运行强大 AI 模型的成本下降了 280 多倍。监控的经济学已经改变。美国美国中央司令部报告称,在“史诗狂怒”行动期间,收到多份威胁报告,涉及对手利用商业位置数据对美军人员实施定位或监视。
Today, a laptop and a data set available for purchase by anyone can get you surprisingly close. AI makes the problem worse. It can sort through enormous datasets and identify patterns that once took an analyst days to piece together. Stanford's AI Index found that the cost of running capable AI models fell more than 280-fold between late 2022 and late 2024. The economics of surveillance have changed. U.S. Central Command has reported receiving multiple threat reports involving adversary use of commercial location data to target or surveil U.S. personnel during Operation Epic Fury.
此类数据价格也可能极其低廉。杜克大学研究人员以每人低至 12 美分的价格,购买到了包含现役军人住址及其子女信息在内的数据。五角大楼终于开始做出回应。9 月 4 日,参议员罗恩·怀登(民主党-俄勒冈州)和众议员帕特·哈里根(共和党-北卡罗来纳州)公布了军方各军种的回复,显示陆军、空军、海军部和特种作战司令部已在政府配发的手机上禁用了移动广告 ID。
The data can also be remarkably cheap. Duke University researchers purchased data about service members that included home addresses and information about their children for as little as 12 cents per person. The Pentagon has finally started responding. On Sept. 4, Sen. Ron Wyden (D-Ore.) and Rep. Pat Harrigan (R-N.C.) released responses from the military services showing that the Army, Air Force, Department of the Navy and Special Operations Command had disabled mobile advertising IDs on government-issued phones.
这堵住了一个真实的漏洞,各军种采取行动值得肯定。但这一举措无法阻止我们今夏发现的那名军官被识别出来。暴露他的不是政府配发的手机,而是他自己口袋里的私人手机,加上公开记录及其家人在网上发布的信息。这一区别至关重要,因为管控政府设备的指导原则,并不能消除个人设备产生的数据。怀登和哈里根在 9 月的信中也提出了同样的担忧,包括带入军事基地的个人手机,仍可能是商业位置数据的重要来源。
That closes a real vulnerability, and the services deserve credit for moving on it. But that would not have stopped us from finding the officer we found this summer. His government phone wasn’t what exposed him — we found him through the phone in his own pocket, along with public records and information his family had posted online. That distinction matters, because government guidance governing government devices does not eliminate the data generated by personal devices. Wyden and Harrigan raised the same concern in their September letter, including the possibility that personal phones carried onto military installations remain an important source of commercial location data.
关闭广告 ID 也无法清除已收集的数据。商业数据集可保存多年的历史记录。对手若要调查某名军官,即便该设置已更改,仍可能重建出该军官很久以前居住、训练、出差和工作的地点。我常被问及,使用一次性手机或注重隐私的运营商是否能解决此问题。两者都有帮助,但都无法让问题彻底消失。
Turning off an advertising ID also does nothing to data that has already been collected. Commercial datasets can preserve years of history. An adversary looking at an officer today may be able to reconstruct where that officer lived, trained, traveled and worked long before the setting was changed. I am often asked whether a burner phone or a privacy-focused carrier solves this. Both can help, but neither makes the problem disappear.
当你把常用应用装到备用手机上、在家和办公室之间携带它、连接熟悉的 Wi-Fi 网络,或者经常把它放在大家都知道属于你的那部手机旁边时,这部备用手机就不再匿名了。完全没有蜂窝服务的手机也存在同样的问题。Wi-Fi 网络、应用、广告技术和其他商业系统持续生成信息,这些信息有助于将设备与个人关联起来。蜂窝运营商只是数据轨迹的一部分。
A burner phone stops being anonymous when you load your normal apps onto it, carry it between your home and office, connect it to familiar Wi-Fi networks, or routinely place it next to the phone everyone already knows belongs to you. The same problem exists with a phone that has no cellular service at all. Wi-Fi networks, apps, advertising technology and other commercial systems continue to generate information that can help connect a device to a person. The cellular carrier is only one part of the data trail.
国会已经开始关注这一更大的问题。众议院通过的 2027 财年《国防授权法案》针对政府所谓的“普遍技术监视”问题。该条款将要求国防部长指定一个组织,负责协调识别监视漏洞,并制定全部门战略、新的培训计划,以及在部内建立永久性正式项目的计划。
Congress has started looking at the larger problem. The House-passed Fiscal Year 2027 National Defense Authorization Act addresses what the government calls ubiquitous technical surveillance. The provision would require the Secretary of Defense to designate an organization responsible for coordinating the identification of surveillance vulnerabilities, along with a department-wide strategy, new training, and a plan for establishing a permanent program of record across the department.
商业监视无法完全归入传统的网络安全、情报、通信或部队防护等类别。不同组织各自掌握问题的一部分,没有任何一个组织掌握全部。这使得零散的修补几乎不可避免。关闭一个标识符,另一个数据源就会变得有用;从一家数据经纪人那里删除一条记录,它可能在别处重现;锁定政府设备,旁边的个人手机可能仍在广播足以识别其主人的信息。
Commercial surveillance doesn't fit neatly inside the traditional boxes of cybersecurity, intelligence, communications or force protection. Different organizations own pieces of the problem. Nobody owns all of it. That makes one-off fixes almost inevitable. Turn off one identifier and another source of data becomes useful. Remove a record from one broker and it can reappear somewhere else. Lock down a government device and the personal phone beside it may still be broadcasting enough information to identify its owner.
国防部需要一个负责的归口部门,全面了解这种暴露情况:确定哪些现役军人及其家属暴露程度最高,评估关于他们的哪些信息已在商业渠道可得,判断哪些可以删除或缓解,并为指挥官提供一种方式,在部署前及部署过程中随时间了解个人的数字暴露程度。
The Department needs an accountable owner responsible for understanding that exposure as a whole: identifying which service members and families are most exposed, assessing what information about them is already commercially available, determining what can be removed or mitigated, and giving commanders a way to understand a person's digital exposure before deployment and as it changes over time.
各军种已经表明,当漏洞明确时,它们能够迅速采取行动。关闭广告 ID 是一个有用的步骤。更棘手的问题是该设置之外的一切:多年积累的现有数据、个人手机、应用程序、公共记录、社交媒体,以及一个旨在将所有这些信息关联起来的产业。我们的对手无需发明新的情报能力来利用这些信息,他们只需购买即可。Erik Wittreich 是 Veilant 公司的首席执行官,也是前美国陆军“绿色贝雷帽”特种部队队员。
The services have already shown that they can move quickly when a vulnerability is clear. Turning off advertising IDs was a useful step. The harder problem is everything that remains outside that setting: years of existing data, personal phones, apps, public records, social media and an industry built to connect all of it. Our adversaries don't need to invent a new intelligence capability to exploit that information. They can buy it. Erik Wittreich is chief executive officer at Veilant and a former Army Green Beret.