- WIADOMOŚCI
- ANALIZA
Trading in data, including soldiers' data. Foreign intelligence services can buy it cheaply
Photo. CyberDefence24/Canva
Foreign intelligence services do not need to steal everything they require to gather intelligence about an adversary’s society. Some information can be purchased from brokers, downloaded from public registers or obtained through the advertising market. Data collected to sell products can subsequently help identify potential recruitment targets, identify military personnel and prepare influence operations. In such an environment, protecting privacy also becomes a counterintelligence task and an element of national security.
A market that cuts the cost of reconnaissance
In an article published in “Small Wars Journal”, Robert Green highlights the vulnerability of a state that protects its secrets while allowing information about its own society to be widely shared.
His central argument is that if an adversary can obtain data useful for intelligence gathering and influence activities at low cost, some of the preparations for an operation are carried out by the commercial market. An expensive cyber intrusion is not always necessary because a ready-made channel of access already exists.
Green links this observation to cognitive warfare, which he describes in his essay as competition over people’s perceptions of reality, judgements and decisions. In his view, large collections of personal data make it possible to divide society into groups and tailor messages to them.
Tools developed by the advertising industry can also serve a state interested in intensifying distrust and conflict. Data then becomes material for preparing operations, much like the intelligence gathering that precedes other intelligence activities.
The most important change here concerns where the threat originates. An intelligence service can obtain information without attacking a protected system if the same data, or data sufficiently useful for its purposes, is held by a broker.
An institution’s security therefore cannot be assessed solely on the resilience of its networks. It is also necessary to examine what can be established about its employees, their families and their wider environment from sources outside the institution.
Public data in the service of foreign intelligence
Green’s starting point is material concerning Chinese interest in American voter data.
The document entitled “US Voter Registration for 6 States” describes how a Chinese entity downloaded data from at least six US states through commercial websites on 14 January 2022. It notes that the information was publicly available and that the actual purpose for which it was obtained remained unknown. At the same time, it points to the possibility of using such resources for influence operations.
The document therefore confirms the acquisition of material potentially useful for operational purposes, but does not prove that a successful campaign against voters was conducted.
We can therefore see that access to such datasets does not necessarily involve breaching the security of a state system. US jurisdictions make certain voter information available under different rules, with varying restrictions and fees.
Obtaining publicly available voter data can pose an intelligence threat even if it does not involve hacking electoral systems. Such information can be used to analyse the composition of the electorate, identify target groups and cross-reference them with data from other sources.
The intelligence value of a dataset may increase when it is combined with other information.
A Government Accountability Office (GAO) report describes how scattered digital traces, when combined, can create profiles revealing information relevant to the security of personnel, their families and military operations.
The document concerns the defence sector but illustrates the general mechanism of aggregation: an individual piece of information may be of limited use, whereas linking it with other information can make it possible to identify a person, their relationships or their activities.
Applying this mechanism to voter registers is an analytical inference — the GAO report itself does not document the use of the Chinese database in question.
The potential uses of personal data are also presented by a Duke University research team in a report published in November 2023.
Its authors indicate that information available from data brokers can help profile military personnel, veterans and people around them, prepare blackmail and target information campaigns.
The researchers demonstrated that such data could be purchased and identified shortcomings in buyer verification. However, they did not prove that a foreign state had used the datasets they purchased in a specific operation.
The study therefore confirms the availability of the resource and describes possible threats, while leaving their actual materialisation as a separate question.
Assessing an operation consequently requires distinguishing between its successive stages:
- obtaining data,
- preparing profiles,
- developing the message,
- reaching the intended audience,
- achieving the intended behavioural change.
See also 
This is an analytical distinction that helps organise the evidence. Possession of a database does not establish the accuracy of the profiles; preparing a message does not prove that it was delivered; and an audience member’s exposure to content does not necessarily mean a change in views, voting decisions or willingness to act. Each successive stage requires additional findings. Without them, it is possible to assess an adversary’s capabilities, but not to attribute a proven success to that adversary.
As FactCheck.org points out in its analysis, obtaining voter registration data is not evidence that electoral infrastructure has been compromised or that votes cast have been altered.
An analysis should therefore examine intelligence gathering about society, influence activities targeting voters, and technical interference with voting and vote counting separately. Data can support the first two types of activity, but mere possession of it confirms neither successful influence nor manipulation of an election result.
Buy what is available, steal what is protected
The economic dimension of covert information acquisition is illustrated by the case of the Chinese company i-Soon.
According to the US Department of Justice, the company carried out hacking operations commissioned by China’s Ministry of State Security and Ministry of Public Security. It also conducted its own operations, subsequently offering the data it obtained to government customers.
In a statement issued in early March 2025 outlining prosecutors’ charges, US authorities reported that i-Soon charged approximately $10,000–75,000 for each successfully compromised email account.
Under this model, a private contractor provides intelligence services with the results of an operation: access to a particular resource and the information obtained.
The existence of commercial databases creates a second supply channel. If the required data can be purchased, commissioning a cyber intrusion may prove unnecessary. An intelligence service can combine purchases, open sources, material from previous leaks and its own operations. The choice of method depends on the availability of information, its quality, cost and risk.
However, this calculation cannot be reduced to a simple comparison of the number of records. A compromised email account belonging to a selected individual may contain plans, relationships and documents that a consumer database containing millions of records cannot provide. Conversely, a cheap dataset may be outdated or riddled with errors.
The market’s advantage lies in reducing the cost of preliminary intelligence gathering and increasing the number of available leads. This does not mean replacing all other intelligence methods.
A soldier as a commercial data record
Empirical confirmation of the availability of such information was provided by the Duke University team. In the previously mentioned report, Justin Sherman and his co-authors described purchases of data concerning active-duty military personnel, veterans and their families. These included health and financial information, as well as information about religious practices. The lowest price obtained was approximately 12 cents per record.
See also 
The researchers contacted sellers using both a .org and a .asia domain. They demonstrated inconsistencies in customer verification procedures, including during purchases made using the .asia domain and a website hosted in Singapore.
The team remained physically in the United States, and its website declared an affiliation with Duke University. The researchers examined whether the foreign nature of the contact infrastructure would prompt sellers to conduct more thorough checks on buyers. The team did not purchase location data, although it noted that such data was available. The study demonstrated that the information could be obtained.
The counterintelligence significance of this finding is nevertheless clear. Information about debt, personal problems, family or interests can help identify individuals for further investigation. Such information does not yet constitute an assessment of susceptibility to recruitment, but it allows the search to be narrowed.
The data market can therefore provide preliminary leads that subsequently require verification through other methods. An adversary gains the ability to begin its work with a ready-made list of potentially interesting individuals.
Advertising as a channel for collecting information
Advertising infrastructure is particularly significant because information can circulate even during the process of determining who should be shown an advertisement.
See also 
In the RTB system, or real-time bidding for advertisements, participants receive data enabling them to assess the value of reaching a particular user. The Mobilewall case demonstrates how this mechanism can be used to build datasets extending beyond the mere delivery of advertisements.
According to a complaint filed by the US Federal Trade Commission (FTC) in December 2024, the company collected and retained information from bid requests even when it did not win the auction.
The FTC stated that between January 2018 and June 2020, the company collected more than 500 million unique advertising identifiers linked to precise location data. The allegations also concerned the sale of sensitive data and inadequate verification of user consent.
A similar problem was described by Joseph Cox in an article published in WIRED, prepared in collaboration with 404 Media and concerning Gravy Analytics.
Analysis of leaked files indicated that some data had been obtained through the advertising ecosystem, potentially without the knowledge of users or app developers. The journalists noted that the origins of individual pieces of information and the involvement of the apps mentioned were not clear in every case.
The security implications are therefore broader: an individual may limit what they publish publicly and still leave digital traces through ordinary phone use. Protection must therefore take account of the entire data ecosystem: from apps and advertising providers to aggregators and downstream recipients. Simply prohibiting the publication of photographs taken at a place of duty addresses only part of the problem.
The European dimension of the threats
The problem also affects Europe. A report by the Irish Council for Civil Liberties (ICCL) described the availability of advertising audience segments associated with the military, public administration, the defence industry and security.
The offers analysed included categories relating to personnel and decision-makers in numerous countries, including Poland.
The ICCL highlighted the threat arising from combining such labels with data on movements, financial circumstances and private life. At the same time, a marketing category should be treated as a classification assigned by the data provider. It does not confirm the employment or position of every individual in a given group. A label suggesting a connection to intelligence services is not a verified list of intelligence officers.
This leads to a practical conclusion for Polish institutions: assessments of personnel exposure should also cover the commercial circulation of information. This applies to the military, public administration, the defence industry and critical infrastructure.
An employee’s family or wider circle may also become an intelligence target if their activities make it possible to establish the employee’s daily routine.
This represents an approach to threat analysis, rather than a claim that all these groups have already been investigated by foreign intelligence services.
Influence on society and pressure on individuals
Green emphasises the possibility of using data to personalise messages. Different groups can be presented with different narratives, with topics selected according to their concerns and experiences.
The objective of an operation may be to deepen distrust, weaken support for a particular policy or discourage action. In this sense, information about audiences helps prepare influence activities.
However, not all influence operations require detailed personal profiles, and profiling alone does not guarantee effectiveness.
Green describes the use of data to tailor messages in influence operations, while the previously mentioned ICCL report demonstrates its usefulness in identifying specific individuals and their susceptibility to pressure.
Comparing these publications makes it possible to distinguish two applications of the same information resources: influencing audience groups and preparing actions against a selected individual.
Not every purchase of data by an intelligence service is therefore an element of cognitive warfare. Information can be used for counterintelligence, economic intelligence gathering, preparing an operational contact or establishing locations.
Only the purpose and manner of use make it possible to determine the nature of an activity. Classifying all information gathering as cognitive warfare would blur the distinctions between intelligence gathering, operational preparation and the influence activity itself.
A secure network does not mean a secure institution
The GAO report frames this problem as a threat to personnel, families and operations. It indicated that scattered digital traces, when combined, can create profiles revealing security-relevant information. It also identified gaps in policies, training and threat assessments within the Department of Defense organisations examined.
The document illustrates an organisational challenge: information concerning a single institution is generated in many places, subject to different rules and ownership arrangements. Some comes from official devices, some from private activities and other information from the operational environment. The threat only becomes apparent when these elements are combined. Responsibility for protection must therefore bring together information and communications technology security, personnel protection, counterintelligence and operational security.
Under this approach, families and subcontractors also require attention because an institution’s boundaries do not coincide with the boundaries of information available about it.
Green presents a separate argument concerning children’s data: information collected from childhood may retain its value when some of those individuals subsequently take up employment in public administration, the military or intelligence services.
This is a prediction of long-term risk and an argument for limiting unnecessary data collection, not a finding of the GAO report or evidence of systematic profiling of future civil servants or military personnel.
Make access harder, and data less useful
The US Data Security Program, in force since 8 April 2025, is an example of treating data transfers as a national security issue.
According to the US Department of Justice, the programme establishes measures that operate like export controls: it prohibits or restricts certain transactions that give designated countries and entities access to US government-related data and large collections of sensitive personal data. It covers data including:
- genomic data,
- biometric data,
- health data,
- financial data,
- location data.
The scope of protection depends on the regulations governing data categories, transactions and recipients.
Green proposes extending this approach by controlling onward resale, verifying recipients, making it harder to download public registers in bulk and strengthening the protection of particularly valuable repositories.
The underlying intention is sound — an adversary should face higher costs when obtaining information. Simply increasing the price will not be sufficient against an intelligence service with a large budget.
Equally important are limiting the amount of data collected, how long it is retained and the ability to identify specific individuals.
Defensive measures should also protect access that enables public scrutiny. Journalists, researchers and citizens need to be able to examine how institutions operate. However, it is worth considering whether the bulk export of data that enables individuals to be profiled requires different safeguards from accessing a single piece of information.
Differentiating access should be based on the nature of the dataset and the actual threat, while maintaining transparency of the rules.
Ultimately, the problem lies in the discrepancy between the market price of information and its potential significance for security.
A record worth a few cents to a seller may, when combined with other data, help identify an individual who has access to important resources.
A state seeking to limit hostile operations must therefore also examine what its own market allows others to establish about its citizens.
Defence against influence begins before a hostile message is published — including wherever data about its future recipients is generated and circulated.









