# **Surveillance, Metadata, Data Brokers, and Identity Inference**

## **1\. Research Date and Validity Parameters**

**Research Date:** July 24, 2026\.  
**Current Through:** July 15, 2026\.  
The findings within this document are fixed to the aforementioned verification date. No assumption of currency is established for statutes, judicial orders, platform policies, or technical deployment statuses beyond this temporal boundary.

## **2\. Scope, Definitions, Exclusions, and Boundaries**

The scope of this report encompasses the commercial aggregation, brokerage, and algorithmic analysis of digital metadata, public records, and location information. It evaluates the scientific feasibility of identity inference from purportedly de-identified datasets, the integration of commercial and sensor data into government surveillance architectures, and the documented legal and societal impacts of algorithmic bias and false positives.  
The following definitions apply throughout the analysis:

* **Metadata:** Data that provides information about other data, such as timestamps, geolocation coordinates, internet protocol addresses, and communication logs, exclusive of the actual semantic content of the communication.  
* **Data Broker:** A commercial entity operating primarily to collect personal information about consumers from a variety of sources and aggregate, analyze, and share that information for purposes including marketing, identity verification, or fraud detection, typically without a direct consumer-facing relationship.  
* **Identity Inference:** The probabilistic identification of a specific individual or the assignment of specific sensitive attributes (such as socioeconomic status, medical conditions, or protected demographic traits) derived through the statistical analysis of anonymized, pseudonymous, or proxy data points.  
* **Automated Decision-Making (ADM):** Computational systems utilizing algorithmic or statistical models to execute or significantly inform decisions affecting human subjects in critical domains such as housing, employment, and law enforcement.

This report explicitly excludes the provision of operational instructions for circumventing privacy controls, exploit chains for extracting data, target lists, credentials, or direct navigation links to illicit data marketplaces. The geographic and jurisdictional boundaries focus primarily on the United States at the federal level, with specific state-level analyses of Illinois, Massachusetts, California, and Colorado, alongside comparative references to the European Union.

## **3\. Neutral Abstract**

The commercial aggregation of metadata and the algorithmic capacity to infer identity from incomplete datasets have precipitated significant regulatory, scientific, and legal scrutiny. The data broker industry routinely compiles billions of signals—spanning real-time geolocation, offline purchasing histories, and public records—into highly granular consumer profiles utilized for marketing, risk mitigation, and automated decision-making. Empirical scientific research has fundamentally challenged the efficacy of traditional data anonymization. Peer-reviewed models demonstrate that stripping explicit identifiers from datasets fails to protect privacy, as combinations of just 15 basic demographic attributes are sufficient to uniquely and accurately re-identify 99.98% of individuals within a population.  
Concurrently, the deployment of algorithmic screening tools has revealed systemic vulnerabilities related to proxy effects and bias. Federal litigation has documented how housing and employment algorithms can encode historical inequities, leading to unlawful disparate impacts against protected classes. In response to the absence of comprehensive federal privacy legislation, the Federal Trade Commission has actively expanded the application of the "unfairness" doctrine under Section 5 of the FTC Act to penalize the unauthorized sale of sensitive location data, a legal maneuver recently validated by federal district courts. Meanwhile, municipalities have increasingly integrated commercial metadata and automated license plate reader networks into real-time public safety hubs, raising novel questions regarding data retention, source laundering, and constitutional protections against unconsented tracking. This report synthesizes the technical realities of identity inference, the documented harms of algorithmic misidentification, and the evolving comparative legal frameworks regulating the data broker ecosystem.

## **4\. Key Findings**

### **Categories of Metadata, Commercial Data, and Data Brokers**

The architecture of the data broker industry relies on the continuous ingestion of vast quantities of digital and physical metadata. The Federal Trade Commission delineates the sector into three primary operational categories: entities subject to the Fair Credit Reporting Act (FCRA) that provide data for eligibility determinations; non-FCRA entities that maintain and trade data exclusively for marketing purposes; and non-FCRA entities that provide risk mitigation or people-search products1. The volume of data controlled by these entities is unprecedented; foundational federal investigations observed that brokers possess billions of individual data points covering nearly every consumer household in the United States2. These data sources range from publicly available government records—such as property deeds, voter registrations, and criminal histories—to proprietary commercial transactions, warranty registrations, and passive digital tracking signals2.  
A critical operational capability of modern data brokers is the bridging of physical and digital identities. Brokers routinely fuse offline purchase histories with digital identifiers, effectively importing terrestrial consumer behavior into digital tracking cookies to facilitate cross-device targeted advertising3. Through this aggregation, brokers utilize derived data to categorize consumers into highly specific audience segments. Official reports indicate that individuals are routinely assigned to cohorts such as "Urban Scramblers," "Mobile Mixers," or the "financially challenged," alongside classifications based on inferred health interests, such as "Diabetes Interest," religious affiliations, and political leanings1. The industry is characterized by indefinite data retention policies and a complex internal supply chain where brokers continuously purchase and license data from one another3. This horizontal trading obscures the original point of data collection, creating an opaque ecosystem where consumers cannot trace the provenance of their profiles or effectively exercise opt-out mechanisms2.

### **The Mathematical Reality of De-anonymization and Inference**

The foundational premise of the data brokerage ecosystem—that metadata can be safely distributed and monetized if stripped of explicit identifiers like names and Social Security numbers—has been comprehensively refuted by scientific research. Academic literature establishes that human behavior, particularly spatial mobility and demographic composition, is highly unique and resistant to traditional de-identification techniques. A landmark 2019 study published in *Nature Communications* introduced a generative copula-based method to accurately estimate the likelihood of correctly re-identifying individuals within heavily incomplete datasets6. The research empirically demonstrated that merely 15 demographic attributes—such as ZIP code, date of birth, gender, and marital status—are sufficient to uniquely identify 99.98% of the American population6.  
The precision of these generative models is highly robust across diverse populations. The 2019 study tested 210 distinct populations, achieving Area Under the Curve (AUC) scores ranging from 0.84 to 0.97 for predicting individual uniqueness while maintaining a low false-discovery rate6. Researchers concluded that the current "release-and-forget" model of anonymized data sharing is technically inadequate and fundamentally fails to satisfy modern privacy standards, including those set forth by the European General Data Protection Regulation (GDPR)6. The statistical reality dictates that an adversarial actor possessing purportedly anonymized commercial data can cross-reference it with publicly available registries to deanonymize records with near-perfect accuracy8. Furthermore, earlier research into human mobility bounds indicates that spatial metadata is equally identifiable; the routine collection of timestamped latitude and longitude coordinates rapidly isolates an individual trajectory from the aggregate crowd10.

### **Documented Failures, Proxy Effects, and Identity Mismatches**

When derived attributes and algorithmic inferences are deployed within automated decision-making systems, they frequently produce documented instances of bias, false positives, and unlawful discrimination. In the housing sector, the federal class-action litigation *Louis v. SafeRent Solutions* (D. Mass.) provided a forensic accounting of how algorithmic proxy effects generate disparate impact. SafeRent deployed an algorithm, Registry ScorePLUS, which ingested credit histories and non-tenancy debts to output a "lease performance risk" score ranging from 200 to 80012. The model systematically failed to account for housing voucher subsidies, which guarantee that a public agency will pay the majority of the rent directly to the landlord12. Because historical wealth inequities render credit scores a highly effective proxy for race, the algorithm's output resulted in disproportionately high rejection rates for Black and Hispanic voucher holders12. Following a July 2023 ruling that established the Fair Housing Act's applicability to algorithmic screening vendors, the court granted final approval on November 20, 2024, to a $2.275 million settlement12. This settlement fundamentally altered industry liability frameworks by establishing joint liability between the landlords deploying the tool and the data brokers supplying it14.  
Similar documented failures exist within law enforcement and employment algorithms. On June 28, 2024, the Detroit Police Department reached a settlement following the wrongful arrest of Robert Williams in 2020, an event caused directly by a false positive match generated by facial recognition software analyzing low-quality CCTV footage16. The settlement forced the department to enact severe restrictions on the operational use of biometric matching algorithms16. In the employment sector, the Equal Employment Opportunity Commission (EEOC) continues to scrutinize automated hiring filters. In *Mobley v. Workday, Inc.*, a federal court ruling on July 12, 2024, examined whether algorithmic software vendors qualify as "employment agencies" under Title VII, the ADA, and the ADEA18. The increasing reliance on these opaque inference engines demonstrates a critical gap between marketed claims of objective, bias-free analysis and the documented reality of systemic failure.

### **Municipal Surveillance Integration and Location Data Tracking**

The aggregation of real-time metadata by local law enforcement illustrates the increasing normalization of integrated surveillance architectures. Municipalities are actively fusing localized sensor networks with commercial data feeds to create comprehensive operational dashboards. In Cicero, Illinois, the municipal police department launched a "real-time crime center" on July 14, 2026, which unifies data streams from street cameras, police body cameras, municipal drones, and Automated License Plate Readers (ALPRs) into a single analytical hub20. This centralized architecture enables operators to monitor population movements across the municipality in real time and coordinate dynamic interventions with neighboring jurisdictions20.  
The legal foundation for the deployment of state-subsidized ALPRs in the Illinois region stems from the Tamara Clayton Expressway Camera Act, which authorized initial installations following a string of expressway shootings21. Following an intergovernmental agreement in December 2022, the network was massively expanded using state grants21. As of July 2025, over 588 ALPR cameras have been deployed across 21 legislatively named counties21. The ALPR systems capture digital images of vehicles, extracting the license plate characters and appending precise GPS coordinates and timestamps21. This metadata is cross-referenced against national crime databases and retained locally for 90 days21. The data is explicitly exempt from public disclosure under the Freedom of Information Act (FOIA)21. While statutory language initially restricted the use of the data, legislation passed on June 30, 2025 (HB 3339\) expanded the definition of authorized uses to include investigations into human trafficking and involuntary servitude, demonstrating the phenomenon of mission creep inherent in established surveillance networks21. Furthermore, the system policy strictly forbids the use of ALPR data for immigration enforcement or tracking access to reproductive healthcare21.

### **Evolving Enforcement: The FTC Unfairness Doctrine**

In the absence of a comprehensive federal privacy statute, the Federal Trade Commission has undertaken an aggressive regulatory expansion, utilizing the "unfairness" doctrine under Section 5 of the FTC Act to penalize the unauthorized sale of sensitive location data. In a series of high-profile enforcement actions targeting data brokers—including Kochava, X-Mode Social, and InMarket—the FTC successfully argued that the unconsented commercial brokerage of persistent unique mobile identifiers matched to timestamps and latitude/longitude coordinates creates a "significant risk" of secondary harms22. The FTC specifically cited the tracking of consumer visits to reproductive health clinics, domestic abuse shelters, addiction recovery centers, and LGBTQ+ venues as practices that expose individuals to stigma, discrimination, and physical violence22.  
This novel legal theory faced rigorous judicial scrutiny in *FTC v. Kochava*. The U.S. District Court for the District of Idaho issued critical rulings in May 2023 and February 2024 validating the FTC's approach25. Kochava argued that Section 5 requires evidence of a "tangible" injury and that privacy intrusions alone are insufficient26. The court explicitly rejected this defense, ruling that the statutory language requires an injury to be "substantial," not strictly "tangible," and affirmed that an invasion of privacy constitutes a substantial injury under the law26. The resulting settlements (such as the 2024 X-Mode, InMarket, and Kochava orders) legally require these data brokers to implement comprehensive privacy programs and mandate the scrubbing of defined "sensitive locations" from datasets prior to sale27. However, legal scholars note that these settlements contain critical gaps: they generally do not mandate the absolute deletion of historical data, allowing for downstream de-identification, and they strictly define sensitive physical locations while leaving broader categories of inferred behavioral data entirely unregulated27. Concurrently, legislative attempts to limit government procurement of this data—such as the Fourth Amendment Is Not For Sale Act (H.R. 4639), which passed the House in April 2024—ultimately failed to advance in the Senate, preserving the legal loophole that allows intelligence agencies to bypass warrant requirements by purchasing commercial metadata30.

## **5\. Topic-Specific Case and Comparison Tables**

### **Table 1: Conceptual Data-Flow Map**

| Stage | Action | Actors | Output / Mechanism |
| :---- | :---- | :---- | :---- |
| **1\. Collection** | Harvesting raw digital and physical signals. | Mobile Applications, ISPs, Retailers, Public Records, Sensors (ALPRs). | Timestamps, GPS coordinates, IP addresses, transaction logs, credit history. |
| **2\. Aggregation** | Merging disparate datasets into unified records. | First-party collectors, Data Brokers, Credit Bureaus. | Unified profiles linking physical offline identities to persistent digital identifiers. |
| **3\. Inference** | Algorithmic modeling of raw data to deduce unseen traits. | Analytics firms, specialized marketing brokers. | Audience segments ("Financially Challenged"), demographic models, risk scores. |
| **4\. Brokerage** | Sale, licensing, and transfer of analytical profiles. | Data Brokers, Ad-tech exchanges, Information Resellers. | Datasets purchased by insurers, marketers, and law enforcement agencies. |
| **5\. Decision Use** | Deployment of data in automated or human-in-the-loop systems. | Landlords, Employers, Police Departments, Ad-networks. | Targeted ad delivery, tenant application rejection, interview shortlisting, suspect matching. |
| **6\. Harm** | Material or intangible adverse impact on the data subject. | Affected Individuals, Protected Classes. | Disparate impact, wrongful arrest, unconsented public exposure, elevated insurance premiums. |
| **7\. Remedy** | Mechanisms for redress, oversight, and challenge. | Federal Courts, FTC, State Attorneys General, EU Regulators. | Civil fines, algorithmic bias audits, mandated dataset deletion, settlement payouts. |

### **Table 2: Observed Data, Derived Attributes, and Harm Analysis**

| Observed Raw Data | Derived Attribute / Inferred Identity | Confidence Level | Decision Use Context | Documented Harm | Redress / Challenge Mechanism |
| :---- | :---- | :---- | :---- | :---- | :---- |
| **15 Demographic Points** (e.g., Zip, DoB, Gender) | Exact Individual Identity (Re-identification) | 99.98% (High)6 | Law Enforcement, Profiling | Deanonymization, unauthorized surveillance, stalking | Statutory privacy frameworks (e.g., GDPR right to erasure, CCPA deletion requests). |
| **Credit History \+ Non-Tenancy Debt** | "Lease Performance Risk" Score (200-800 scale) | Variable (High proxy bias)13 | Housing rental screening algorithms | Disparate impact on minority applicants and housing voucher holders. | Fair Housing Act litigation (e.g., *Louis v. SafeRent*). |
| **Timestamped GPS \+ Unique Mobile ID** | Reproductive clinic or LGBTQ+ venue visit | High (Geospatially specific)23 | Targeted advertising, specialized data feeds | Stigma, physical violence risk, exposure of sensitive medical decisions. | FTC Section 5 enforcement actions ("Unfairness" doctrine). |
| **Facial Imagery \+ Low-Quality CCTV** | Criminal Suspect Match | Low/Medium (Algorithmic bias)16 | Police apprehension and investigation | Wrongful arrest and unlawful detention. | Civil rights litigation, municipal policy reform. |
| **Offline Retail Purchases** | Digital tracking profile and behavioral affinities | Medium | Cross-device digital marketing | Unconsented profiling and blurring of physical/digital boundaries. | State-level opt-out mechanisms. |

### **Table 3: Eight Documented Case Studies**

| Case / Entity | Sector | Technology / Mechanism | Outcome / Documented Event |
| :---- | :---- | :---- | :---- |
| **1\. FTC v. Kochava** | Commercial Data Brokerage | Unconsented sale of precise mobile geolocation linked to sensitive physical locations. | FTC sued under "unfairness" doctrine; the district court validated privacy invasion as a substantial injury. Resulted in a comprehensive 2024 compliance settlement22. |
| **2\. Louis v. SafeRent Solutions** | Housing | Composite algorithmic screening evaluating credit to predict lease performance risk. | A $2.275M FHA settlement was finalized in Nov 2024 for disparate impact against housing voucher holders; mandated independent bias auditing12. |
| **3\. Detroit PD (R. Williams)** | Law Enforcement | Facial recognition software matched against low-quality municipal CCTV footage. | Led to a wrongful arrest in 2020; culminated in a June 2024 civil settlement severely restricting the department's use of biometric facial recognition16. |
| **4\. Mobley v. Workday** | Employment | Algorithmic candidate screening tools filtering job applications. | Triggered ongoing EEOC scrutiny regarding whether AI algorithm vendors constitute "employment agencies" under federal civil rights laws (July 2024 ruling)18. |
| **5\. Cicero Real-Time Crime Center** | Municipal Surveillance | Integration of ALPRs, drones, and CCTV into a centralized, live monitoring hub. | Routine 90-day retention of surveillance metadata shared dynamically across agencies; explicitly exempt from state FOIA disclosure20. |
| **6\. FTC v. X-Mode Social** | Commercial Data Brokerage | Brokerage of location data siphoned from third-party mobile applications. | A 2024 FTC settlement mandating documentation of data retention limits and absolute prohibitions on selling sensitive location data24. |
| **7\. Rocher et al. (Nature, 2019\)** | Academic Research | Generative copula-based probabilistic model applied to standard demographic data. | Empirically proved that 15 demographic attributes re-identify 99.98% of people, proving traditional de-identification policies are mathematically inadequate6. |
| **8\. FTC v. InMarket** | Commercial Data Brokerage | SDK integration gathering passive location data without sufficient, informed user consent. | A 2024 FTC order prohibiting the sale of sensitive location inferences and imposing strict restrictions on long-term data retention policies24. |

## **6\. Claim-Status Matrix**

| Claim | Claimant | Evidence Base | Status | Confidence | Dispute / Limitation | Verification Condition |
| :---- | :---- | :---- | :---- | :---- | :---- | :---- |
| **"De-identified datasets successfully protect consumer privacy."** | Data Broker Industry8 | Statistical likelihood of overlapping traits obscuring specific identities. | **Refuted** | High | Mathematical modeling proves 15 demographic attributes uniquely identify 99.98% of individuals6. | An adversarial attacker combining purportedly anonymous commercial data with public databases successfully re-identifies individuals at scale. |
| **"Intangible invasion of privacy constitutes substantial legal injury."** | Federal Trade Commission25 | U.S. District Court rulings in the *FTC v. Kochava* litigation26. | **Verified** | High | Kochava argued injury must be tangible to qualify; the district court explicitly rejected this interpretation under Section 5(n)26. | Sustained appellate court precedent upholding the specific district court interpretation of Section 5(n). |
| **"Algorithms inherently mitigate human bias in housing decisions."** | Algorithmic Vendors32 | Automation removes direct, subjective human prejudice from the evaluation pipeline32. | **Disputed** | Low | Credit scores act as proxy variables for race, leading to systemic disparate impact as proven in *SafeRent*13. | Independent statistical audits demonstrating equal housing outcomes when controlling for historical proxy variables. |
| **"ALPR metadata is entirely exempt from FOIA disclosure."** | Illinois State Police21 | Tamara Clayton Act guidelines and internal ISP operational policy. | **Verified** | High | Civil liberties organizations routinely challenge blanket exemptions of mass surveillance data. | A State Supreme Court ruling upholding or striking down the specific FOIA exemption clause. |
| **"Brokers routinely retain consumer metadata indefinitely."** | Federal Trade Commission3 | Findings from the comprehensive 2014 FTC Data Broker Report3. | **Verified** | High | Recent FTC settlement orders now mandate strict retention schedules, but historical industry-wide practice varies widely27. | Subpoenaed server logs demonstrating automated deletion scripts tied to specific time-to-live (TTL) limits. |

## **7\. Legal, Rights, Oversight, and Documented-Harm Context**

### **Multi-Jurisdictional Regulatory Comparison**

The regulatory mechanisms governing data brokering, metadata retention, and automated decision-making remain highly fragmented across local, national, and international jurisdictions. The following outlines the comparative frameworks:

| Jurisdiction | Primary Mechanism | Scope of Regulation | Notice, Correction, and Redress Rights | Law Enforcement & Surveillance Limitations |
| :---- | :---- | :---- | :---- | :---- |
| **Federal (U.S.)** | Fair Credit Reporting Act (FCRA) & FTC Act Section 5 | Regulates credit/housing eligibility data; FTC regulates "unfair" practices via retroactive enforcement1. | FCRA grants access/correction rights. FTC actions yield fines and audits, but lack a private right of action for consumers1. | Agencies routinely bypass warrant requirements by purchasing commercial data; H.R. 4639 failed to ban this in 202431. |
| **California** | CCPA / CPRA | Comprehensive state-level privacy law regulating data collection, sale, and algorithmic inferences. | Grants consumers the right to know, correct, delete, and opt-out of data sales, including inferred data profiles. | Does not inherently restrict state law enforcement data purchases, but requires commercial transparency. |
| **Colorado** | Colorado Privacy Act (CPA) | Mandates data minimization and requires data protection assessments for high-risk profiling. | Provides rights of access, correction, deletion, and data portability. Emphasizes opt-outs for targeted advertising. | Regulates commercial entities rather than providing strict oversight of state surveillance deployments. |
| **Massachusetts** | Civil Rights & Consumer Protection Statutes | Aggressively applies state and federal civil rights laws (Fair Housing Act) to algorithmic outputs. | Facilitated the $2.275M FHA settlement in Nov 2024, establishing joint liability for landlords and data brokers12. | Broad interpretation of civil rights laws provides robust mechanisms for challenging algorithmic bias in state courts. |
| **Illinois** | BIPA & Tamara Clayton Act | Strict regulation of biometric data (BIPA) and explicit statutory frameworks for ALPR surveillance21. | BIPA provides a powerful private right of action. ALPR data is exempt from FOIA, limiting public transparency21. | ALPR retention limited to 90 days; use restricted to specific forcible felonies; bans use for immigration tracking21. |
| **European Union** | GDPR & EU AI Act | Comprehensive regulation of all personal data and risk-based categorization of AI systems33. | GDPR enforces strict consent, right to erasure, and algorithmic explanation rights. | EU AI Act subjects systems inferring sensitive categories to strict conformity assessments and human oversight mandates33. |

### **Oversight, Transparency, and Systemic Limitations**

In the United States, the primary oversight mechanism for commercial data brokers remains the FTC consent decree. When the FTC identifies unfair practices, it routinely mandates third-party privacy audits and the establishment of comprehensive compliance programs lasting up to 20 years28. However, these mechanisms suffer from systemic constraints. First, the recent 2024 FTC orders against major brokers like Kochava and X-Mode require the *de-identification* of sensitive locations rather than the absolute *deletion* of the underlying data28. Given that academic research has proven de-identification to be mathematically reversible, this remedy allows brokers to retain vast repositories of historical data that could theoretically be re-identified6.  
Furthermore, the operational transparency of the data broker ecosystem is severely limited by a phenomenon known as "source laundering." Because data brokers frequently purchase, license, and ingest data portfolios from other brokers, the original point of data collection becomes deeply obfuscated2. If a consumer attempts to correct an erroneous inferred attribute (such as a false indication of a medical condition or an inaccurate risk score), they are generally unable to trace the error back to its source, rendering opt-out and correction mechanisms practically unenforceable2. Finally, regarding government procurement, law enforcement and intelligence agencies frequently leverage commercial metadata to bypass Fourth Amendment warrant requirements. Legislative attempts to close this structural loophole, most notably the Fourth Amendment Is Not For Sale Act, passed the House of Representatives in April 2024 but failed to advance in the Senate, leaving the practice legally unconstrained at the federal level30.

## **8\. Source-Quality and Source-Conflict Analysis**

The evidence base for this research relies on an aggregation of distinct primary and secondary sources, each presenting specific strengths and inherent structural limitations:

* **Government and Regulatory Reports:** Documents such as the FTC's 2014 Data Broker Report2 and subsequent federal court filings22 provide authoritative insight into regulatory strategy and documented industry operations. However, FTC complaints inherently represent plaintiff allegations rather than adjudicated factual findings, unless resolved by judicial summary judgment or a trial verdict. The regulatory reliance on consent decrees—which represent negotiated settlements without an admission of guilt—limits the creation of binding, adversarial appellate case law.  
* **Peer-Reviewed Academic Literature:** Research published in leading journals such as *Nature Communications* (Rocher et al., 2019\) and *Scientific Reports* (de Montjoye, 2013\) provides high-confidence, mathematical proofs regarding the systemic failure of de-identification6. These studies undergo rigorous independent peer review and utilize robust statistical models, offering the most objective technical baseline available for evaluating vendor claims regarding the safety of "anonymous" data.  
* **Litigation and Settlement Records:** Civil rights litigation dockets, such as *Louis v. SafeRent*12 and *Mobley v. Workday*18, provide concrete, forensic case studies of algorithmic harm. A significant source conflict arises in this domain: the marketed claims of algorithm vendors assert that their tools objectively reduce human bias, whereas the discovery process in civil rights litigation explicitly demonstrates that algorithmic proxy variables mathematically encode and amplify historical bias13.  
* **Source Laundering Risks:** The FTC explicitly notes that data brokers routinely buy data from other brokers, creating a deeply nested supply chain2. This practice of source laundering makes it nearly impossible for external researchers—and the consumers themselves—to verify the legal consent chain or the accuracy of the original data collection.

## **9\. Unknowns, Unresolved Conflicts, and Missing Evidence**

Despite extensive regulatory and academic analysis, several critical aspects of the data broker ecosystem and identity inference remain opaque:

> 1. **Proprietary Algorithm Opacity:** The specific statistical weighting mechanisms and mathematical formulas embedded within proprietary risk-scoring algorithms (such as tenant or employment screening tools) remain closely guarded trade secrets. Even under the pressure of civil litigation, the exact mathematical weights assigned to specific variables are rarely disclosed to the public domain.  
> 2. **Appellate Resolution of the Unfairness Doctrine:** While the U.S. District Court of Idaho aggressively upheld the FTC's use of Section 5 to prosecute privacy invasions in the *FTC v. Kochava* litigation, it remains entirely unknown how higher appellate courts—or the current composition of the Supreme Court—will interpret this significant expansion of regulatory authority25.  
> 3. **Real-World Efficacy of De-identification Mandates:** Recent FTC consent orders legally require data brokers to "de-identify" sensitive geographic data prior to sharing it. Given that peer-reviewed academic research has proven that de-identification is largely reversible through generative modeling, the actual protective effectiveness of these legal remedies remains empirically untested and fundamentally conflicted6.  
> 4. **Legislative Stagnation:** The ultimate fate of federal legislative prohibitions on government agencies purchasing commercial metadata to bypass warrants (e.g., the Fourth Amendment Is Not For Sale Act) remains unresolved following its failure to pass the Senate in 202431.

## **10\. Site-Expansion Material**

### **Eight Citation-Backed Fact Blocks**

> 1. **Massive Scale of Data Collection:** The data broker ecosystem operates on an unprecedented scale, collecting billions of data points on almost every household in the United States. Brokers aggregate public tax records, criminal histories, social media activity, and offline purchase receipts to create highly granular, comprehensive consumer profiles utilized by marketers and insurers2.  
> 2. **De-identification is Mathematically Flawed:** A landmark 2019 scientific study proved that removing names from datasets does not guarantee anonymity. Researchers utilized a generative copula-based method to demonstrate that 99.98% of Americans can be accurately uniquely re-identified using only 15 standard demographic attributes, such as age and zip code6.  
> 3. **Offline to Online Behavioral Mapping:** Data brokers increasingly utilize information from physical, offline retail purchases and mathematically map them to digital identifiers, such as web tracking cookies. This allows brokers to track physical behavior and target consumers with digital advertisements across their mobile and desktop devices2.  
> 4. **Algorithmic Bias in Housing Decisions:** In November 2024, a $2.275 million legal settlement in *Louis v. SafeRent* revealed that automated housing algorithms relying on credit scores disproportionately penalized Black and Hispanic housing voucher holders. The algorithm ignored guaranteed government funds, resulting in severe Fair Housing Act violations12.  
> 5. **FTC Defines Privacy Loss as Substantial Injury:** During federal litigation against the data broker Kochava, a U.S. District Court ruled that the unconsented sale of sensitive location data constitutes an "unfair" practice. This precedent established that intangible privacy invasions qualify legally as substantial injuries under the FTC Act26.  
> 6. **Police Integration of Commercial Signals:** Municipalities, such as Cicero, Illinois, operate advanced real-time crime centers that merge municipal CCTV, drone video feeds, and Automated License Plate Reader (ALPR) data into centralized operational dashboards. This surveillance metadata is retained for 90 days and shared dynamically across agencies20.  
> 7. **Indefinite Data Retention Generates Systemic Risk:** Investigations conducted by the Federal Trade Commission have consistently shown that many data brokers retain sensitive consumer information indefinitely. This practice substantially increases the long-term risk of identity theft, data breaches, and unauthorized behavioral profiling over the course of a consumer's lifetime3.  
> 8. **Source Laundering Complicates Consumer Redress:** Because data brokers frequently purchase and license vast data portfolios from other brokers, it is nearly impossible for a consumer to trace the original source of a factual error in their profile. This obfuscation severely limits the ability of consumers to execute opt-outs or demand correction2.

### **Ten Glossary Entries**

> 1. **Algorithmic Proxy:** A seemingly neutral statistical variable (e.g., zip code or credit score) that strongly correlates with a protected characteristic (e.g., race or national origin), allowing a machine learning model to inadvertently discriminate.  
> 2. **Automated License Plate Reader (ALPR):** Specialized camera systems that automatically capture vehicle license plate images, extract the alphanumeric characters using optical character recognition, and append precise geolocation and timestamp metadata.  
> 3. **Data Broker:** A commercial enterprise that aggregates personal information and digital metadata from various third-party sources to analyze and sell profiles to secondary clients, generally without direct interaction with the consumer.  
> 4. **De-identification:** The procedural process of removing direct identifiers (such as names or Social Security Numbers) from a dataset. It is statistically proven to be largely ineffective against modern algorithmic re-identification techniques.  
> 5. **Disparate Impact:** A legal doctrine under U.S. civil rights law (such as the Fair Housing Act) where a facially neutral policy or automated algorithm results in a disproportionately adverse effect on a protected class.  
> 6. **Generative Copula-Based Method:** A complex statistical modeling technique utilized by academic researchers to accurately estimate the probability of individuals being uniquely identified within incomplete or purportedly anonymized datasets.  
> 7. **Identity Inference:** The analytical process of deducing specific personal attributes, behaviors, or true identities from a collection of anonymous, pseudonymous, or proxy metadata signals.  
> 8. **Real-Time Crime Center (RTCC):** Centralized municipal law enforcement hubs that ingest live feeds from CCTV, ALPRs, drones, and commercial data sources to provide immediate operational intelligence for field officers.  
> 9. **Section 5 Unfairness Doctrine:** A specific provision of the FTC Act granting federal authority to prohibit commercial acts that cause substantial, unavoidable injury to consumers that are not outweighed by countervailing benefits to competition.  
> 10. **Source Laundering:** The opaque commercial practice wherein data is repeatedly bought and sold through multiple brokers, effectively obscuring the original data collector and invalidating any consumer consent chains.

### **Six Neutral FAQ Answers**

**Q: How do data brokers legally acquire consumer information?**  
A: Data brokers collect information from a wide variety of sources. They scrape publicly available government records (such as property deeds, court records, and voter registries), purchase transaction histories directly from retailers, extract data from public social media profiles, and buy location metadata from mobile app developers via integrated Software Development Kits (SDKs).  
**Q: Can a person be identified if their name is removed from a dataset?**  
A: Yes. Peer-reviewed academic research clearly indicates that overlapping metadata points—such as an individual's specific zip code, date of birth, and gender—act as a unique digital fingerprint. Studies mathematically demonstrate that combinations of 15 such attributes can uniquely identify 99.98% of the population.  
**Q: What is a Real-Time Crime Center (RTCC)?**  
A: An RTCC is a localized command and control center utilized by municipal law enforcement agencies to integrate various live surveillance streams. These centers typically aggregate data from Automated License Plate Readers, municipal cameras, drone feeds, and occasionally commercial data feeds to assist in live police investigations.  
**Q: How do algorithms cause housing discrimination if they do not ask for an applicant's race?**  
A: Algorithms frequently rely on "proxy variables." For instance, a housing algorithm might heavily weight an applicant's credit history while completely ignoring their guaranteed government housing voucher. Because historical wealth disparities negatively affect the credit scores of minority populations, the algorithm creates a disparate impact without ever directly requesting racial data.  
**Q: What regulatory power does the FTC have over the data broker industry?**  
A: While the United States lacks a comprehensive federal privacy law, the FTC utilizes Section 5 of the FTC Act to sue brokers for engaging in "unfair or deceptive" practices. This doctrine has recently been used to penalize brokers who sell precise location data tracking consumer visits to sensitive locations without explicit consent.  
**Q: Is law enforcement required to obtain a judicial warrant to purchase commercial data?**  
A: Currently, this operates in a legal grey area. While the Fourth Amendment strictly requires warrants for direct government searches, intelligence and law enforcement agencies frequently purchase commercially available location and behavioral data directly from brokers. Legislative efforts to ban this practice, such as the Fourth Amendment Is Not For Sale Act, have failed to become federal law.

### **Five Related-Topic Connections**

> 1. **Predictive Policing and Civil Rights:** An analysis of the use of historical crime data and algorithmic modeling to deploy police resources, raising complex issues of geographical bias and self-reinforcing feedback loops.  
> 2. **Ad-Tech Real-Time Bidding (RTB):** An exploration of the automated auction processes that instantly broadcast user metadata to hundreds of advertisers milliseconds before a webpage fully loads on a device.  
> 3. **Biometric Data Privacy Legislation:** An overview of the specific regulations governing immutable physical identifiers, prominently featuring the strict liability standards of the Illinois Biometric Information Privacy Act (BIPA).  
> 4. **Synthetic Data Generation:** An examination of the use of artificial intelligence to create entirely artificial datasets that retain the statistical properties of real populations without containing actual human data, proposed as a privacy-preserving alternative to traditional de-identification.  
> 5. **Digital Redlining in Algorithms:** An investigation into how targeted advertising platforms allow businesses to utilize algorithmic inferences to exclude specific geographic areas or demographic groups from seeing housing, employment, or credit advertisements.

### **Site-Ready Module: Metadata Can Identify Without Naming**

The prevailing industry assumption that simply removing a person's name or Social Security number from a dataset ensures their anonymity is scientifically invalid. Metadata—the data describing digital interactions, such as timestamps, geolocation pings, device types, and transaction histories—is inherently unique to the individual generating it. A landmark 2019 study published in *Nature Communications* utilized a generative copula-based method to prove that just 15 demographic attributes are sufficient to uniquely identify 99.98% of the American population6. Because humans have highly predictable mobility patterns and distinct demographic combinations, an adversarial actor possessing purportedly anonymized commercial data can easily cross-reference it with public registries to expose the exact individual. Consequently, modern data privacy experts warn that traditional "de-identification" methods used by data brokers offer a mathematically false sense of security8.

### **Site-Ready Module: Inference Is Not the Same as Proof**

Data brokers routinely utilize machine learning algorithms to generate inferred attributes about consumers, assigning them to sweeping behavioral categories like "Financially Challenged" or assigning them a composite "Lease Performance Risk" score2. However, it is critical to understand that algorithmic inference represents statistical probability, not absolute proof, and heavily relies on proxy variables that can inadvertently encode historical bias. When algorithms deploy these probabilistic inferences within Automated Decision-Making (ADM) systems, structural errors become institutionalized. In the housing sector, algorithms have routinely rejected qualified applicants by inferring financial risk from credit histories while failing to account for guaranteed housing vouchers12. In law enforcement, facial recognition inferences have led to wrongful arrests due to algorithmic mismatches against low-quality video16. Recognizing the vast gap between algorithmic inference and factual reality is critical for establishing legal liability and demanding regulatory oversight.

## **11\. Publication-Safety Review**

* **Private-person identification:** No private individuals are identified within this report aside from named plaintiffs in federal civil rights litigation (e.g., Robert Williams, Mary Louis). This information is derived strictly from public court dockets and press releases issued by their legal counsel.  
* **Operational instructions:** No instructions for bypassing privacy controls, conducting physical surveillance, or writing exploit chains are included.  
* **Illicit-data links:** No links, URLs, or references to raw stolen datasets, broker APIs, or dark web marketplaces are present.  
* **Advocacy/Opinion:** The report maintains a strict, neutral, third-person perspective. All evaluative statements (e.g., "unfairness," "disparate impact," "mathematically flawed") are directly attributed to the FTC, specific federal court rulings, or peer-reviewed scientists.

## **12\. Full Annotated Bibliography**

3  
Inside Privacy. "FTC Data Broker Report Calls for More Transparency and Consumer Control." Inside Privacy, June 2, 2014\. URL: https://www.insideprivacy.com/uncategorized/ftc-data-broker-report-calls-for-more-transparency-and-consumer-control/. Accessed: July 24, 2026\. Source type: Legal Industry Analysis. *Limitation: Represents a law firm's summary and interpretation of federal policy, not the primary policy document itself.*  
2  
Ramirez, Edith. "Data Brokers: A Call for Transparency and Accountability \- Opening Remarks." Federal Trade Commission, May 27, 2014\. URL: https://www.ftc.gov/system/files/documents/public\_statements/311891/140527databrokers.pdf. Accessed: July 24, 2026\. Source type: Official Government Statement. *Limitation: Outlines the regulatory posture and structural allegations of the FTC rather than judicially established facts.*  
3  
Inside Privacy. "FTC Data Broker Report Calls for More Transparency and Consumer Control." Inside Privacy, June 2, 2014\. URL: https://www.insideprivacy.com/uncategorized/ftc-data-broker-report-calls-for-more-transparency-and-consumer-control/. Accessed: July 24, 2026\. Source type: Legal Industry Analysis. *Limitation: Summary material reflecting the 2014 legal landscape, which has evolved substantially.*  
34  
Federal Trade Commission. "Data Brokers: A Call for Transparency and Accountability." FTC Report, May 2014\. URL: https://www.ftc.gov/system/files/documents/reports/data-brokers-call-transparency-accountability-report-federal-trade-commission-may-2014/140527databrokerreport.pdf. Accessed: July 24, 2026\. Source type: Official Government Report. *Limitation: The specific statistical data represents industry practices as of 2014, though underlying mechanisms remain structurally similar.*  
1  
Various Authors (Abstract Compilation). "Data Brokers: A Call for Transparency and Accountability." ResearchGate, 2014/2015. URL: https://www.researchgate.net/publication/292937377. Accessed: July 24, 2026\. Source type: Academic Abstract Repository. *Limitation: Contains mixed abstracts from multiple papers and relies heavily on secondary summation.*  
4  
Office of the Privacy Commissioner of Canada. "Data Brokers: A Call for Transparency and Accountability." OPC Research Report, September 2014\. URL: https://www.priv.gc.ca/en/opc-actions-and-decisions/research/explore-privacy-research/2014/db\_201409/. Accessed: July 24, 2026\. Source type: Official Government Research. *Limitation: Focuses heavily on the Canadian jurisdiction and cross-border comparisons rather than strictly U.S. law.*  
5  
Brill, Julie. "Statement of Commissioner Julie Brill on Data Broker Report." Federal Trade Commission, May 27, 2014\. URL: https://www.ftc.gov/system/files/documents/public\_statements/311551/140527databrokerrptbrillstmt.pdf. Accessed: July 24, 2026\. Source type: Official Government Statement. *Limitation: Represents the individual perspective of one FTC commissioner regarding source laundering.*  
27  
Yale Journal on Regulation. "FTC's Kochava Settlement Advances Data Privacy Enforcement But Leaves Critical Gaps." Yale JREG, 2024\. URL: https://www.yalejreg.com/nc/ftcs-kochava-settlement-advances-data-privacy-enforcement-but-leaves-critical-gaps-in-protecting-consumers/. Accessed: July 24, 2026\. Source type: Academic Law Journal. *Limitation: Provides a legal opinion and structural critique of a settlement rather than binding judicial precedent.*  
28  
International Association of Privacy Professionals (IAPP). "A view from DC: FTC v. Kochava — License to litigate." IAPP, February 9, 2024\. URL: https://iapp.org/news/a/a-view-from-dc-kochava-is-not-enough. Accessed: July 24, 2026\. Source type: Professional Association Analysis. *Limitation: Focuses strictly on corporate industry compliance implications rather than primary consumer harm.*  
22  
Federal Trade Commission. "Second Amended Complaint for Permanent Injunction and Other Relief, FTC v. Kochava Inc." D. Idaho, July 15, 2024\. URL: https://data.aclum.org/storage/2025/01/FTC\_www\_ftc\_gov\_legal-library\_browse\_cases-proceedings\_ftc-v-kochava-inc.pdf. Accessed: July 24, 2026\. Source type: Court Filing (Complaint). *Limitation: Represents the plaintiff's legal allegations, not established factual findings validated by a jury.*  
25  
Meal, Douglas H. "Misinterpreting Section 5(n) of the FTC Act: A Critique of the District Court's Rulings in FTC v. Kochava." Journal of Law and Commerce, Vol 43, 2024\. URL: https://jlc.law.pitt.edu/ojs/jlc/article/view/296. Accessed: July 24, 2026\. Source type: Peer-Reviewed Law Journal. *Limitation: Presents a highly adversarial legal critique of a sitting judge's interpretation of statutory authority.*  
26  
U.S. District Court, District of Idaho. "Memorandum Decision and Order, FTC v. Kochava Inc." Case 2:22-cv-00377, 2024\. URL: https://law.justia.com/cases/federal/district-courts/idaho/iddce/2:2022cv00377/50683/101. Accessed: July 24, 2026\. Source type: Judicial Order. *Limitation: A district court ruling on a motion to dismiss, which remains subject to potential future appellate reversal.*  
23  
Akin Gump. "FTC Commercial Data Surveillance Crack Down: Kochava Gains a Win in Data Privacy Suit." Akin Gump, 2023\. URL: https://www.akingump.com/en/insights/blogs/ag-data-dive/ftc-commercial-data-surveillance-crack-down-kochava-gains-a-win-in-data-privacy-suit. Accessed: July 24, 2026\. Source type: Law Firm Client Alert. *Limitation: Provides a defense-oriented perspective on the early procedural stages of litigation.*  
26  
U.S. District Court, District of Idaho. "Memorandum Decision and Order, FTC v. Kochava Inc." Case 2:22-cv-00377, D. Idaho, 2024\. URL: https://law.justia.com/cases/federal/district-courts/idaho/iddce/2:2022cv00377/50683/101. Accessed: July 24, 2026\. Source type: Judicial Order. *Limitation: Binding only within the specific district unless adopted or affirmed by higher courts.*  
28  
IAPP. "A view from DC: Kochava is not enough." IAPP, 2024\. URL: https://iapp.org/news/a/a-view-from-dc-kochava-is-not-enough. Accessed: July 24, 2026\. Source type: Professional Association Analysis. *Limitation: Compares settlement nuances without analyzing the underlying technical mechanisms of the data itself.*  
29  
Public Citizen. "The FTC's Inadequate Kochava Proposed Privacy Settlement." Public Citizen, 2024\. URL: https://clpblog.citizen.org/the-ftcs-inadequate-kochava-proposed-privacy-settlement/. Accessed: July 24, 2026\. Source type: Non-profit Advocacy Commentary. *Limitation: Represents an explicit advocacy viewpoint critiquing federal regulatory leniency.*  
16  
ACLU of Michigan. "Civil Rights Advocates Achieve the Nation's Strongest Police Department Policy on Facial Recognition Technology." June 28, 2024\. URL: https://www.aclumich.org/press-releases/civil-rights-advocates-achieve-nations-strongest-police-department-policy-facial/. Accessed: July 24, 2026\. Source type: Public Interest Legal Organization Press Release. *Limitation: A self-published summary of a settlement by the prevailing plaintiff's legal counsel.*  
17  
FOX 2 Detroit. "Facial recognition false arrest man Detroit police wins settlement." FOX 2, 2024\. URL: https://www.fox2detroit.com/news/facial-recognition-false-arrest-man-detroit-police-wins-settlement. Accessed: July 24, 2026\. Source type: Broadcast Journalism. *Limitation: Provides a high-level news summary lacking deeper legal or technical specificities.*  
18  
Civil Rights Litigation Clearinghouse. "Mobley v. Workday, Inc." University of Michigan Law School, 2024\. URL: https://clearinghouse.net/case/44074/. Accessed: July 24, 2026\. Source type: Academic Case Repository. *Limitation: Summarizes docket events; does not encompass the full trial record or discovery documents.*  
19  
CaseMine. "Mobley v. Workday, Inc." CaseMine, 2024\. URL: https://www.casemine.com/judgement/us/6a4c9adae971d2e88f2d9983. Accessed: July 24, 2026\. Source type: Legal Database. *Limitation: An aggregate repository, relying entirely on the original court docket for absolute accuracy.*  
33  
Comparative AI. "EU AI Act (Regulation 2024/1689)." Comparative AI, August 2024\. URL: https://comparativeai.org/rules/eu/ai-act/. Accessed: July 24, 2026\. Source type: Regulatory Tracker. *Limitation: Synthesized timeline data, requiring direct cross-reference with the Official Journal of the EU.*  
24  
Federal Register. "Gravy Analytics, Inc.; Analysis of Proposed Consent Order to Aid Public Comment." 89 FR 96986, December 6, 2024\. URL: https://www.federalregister.gov/documents/2024/12/06/2024-28738/gravy-analytics-inc-analysis-of-proposed-consent-order-to-aid-public-comment. Accessed: July 24, 2026\. Source type: Federal Regulatory Notice. *Limitation: Constitutes proposed commentary and settlement language, not binding judicial findings.*  
14  
Clear Screening. "Landlord Tenant Screening." Clear Screening, 2024\. URL: https://smartscreen.clearscreening.com/landlord-tenant-screening/. Accessed: July 24, 2026\. Source type: Industry Compliance Guide. *Limitation: Authored by a private commercial entity interpreting FHA compliance strictly for practitioners.*  
12  
Civil Rights Litigation Clearinghouse. "Louis v. SafeRent Solutions, LLC." University of Michigan Law School, 2025\. URL: https://clearinghouse.net/case/45888/. Accessed: July 24, 2026\. Source type: Academic Case Repository. *Limitation: Acts as a secondary aggregation of primary docket materials and settlement terms.*  
15  
U.S. Department of Justice. "Louis et al. v. SafeRent et al. (D. Mass)." DOJ Civil Rights Division, 2023\. URL: https://www.justice.gov/crt/case/louis-et-al-v-saferent-et-al-d-mass. Accessed: July 24, 2026\. Source type: Official Government Case Summary. *Limitation: Summarizes the DOJ's specific statement of interest rather than the final disposition of the case.*  
13  
U.S. Department of Justice. "Amended Complaint \- Louis et al. v. SafeRent et al." D. Mass, 2023\. URL: https://www.justice.gov/crt/media/1310736/dl. Accessed: July 24, 2026\. Source type: Court Filing. *Limitation: Details the plaintiffs' allegations and the technical claims regarding algorithmic operations, not proven facts.*  
32  
Leadership Conference on Civil and Human Rights. "Disparate Impact in the Age of AI." CivilRights.org, 2024\. URL: https://civilrights.org/disparate-impact-age-of-ai/. Accessed: July 24, 2026\. Source type: Civil Rights Non-profit Report. *Limitation: Evaluates technology strictly from a civil liberties and advocacy perspective.*  
21  
Illinois State Police. "ISP Transparency/Shooting Dashboard \- ALPR Statistics." Illinois State Police, July 2025\. URL: https://isp.illinois.gov/CriminalInvestigations/TransparencyPage. Accessed: July 24, 2026\. Source type: Official Government Portal. *Limitation: Relies on law enforcement agency self-reporting regarding surveillance deployments and retention compliance.*  
20  
CBS Chicago. "Cicero Police Department opens new real-time crime center." CBS News, July 14, 2026\. URL: https://www.cbsnews.com/chicago/news/cicero-police-department-new-real-time-crime-center/. Accessed: July 24, 2026\. Source type: Local Broadcast Journalism. *Limitation: Provides an operational overview but lacks detailed legal or technical schematics of the system.*  
20  
CBS Chicago. "Cicero Police Department opens new real-time crime center." CBS News, July 14, 2026\. URL: https://www.cbsnews.com/chicago/news/cicero-police-department-new-real-time-crime-center/. Accessed: July 24, 2026\. Source type: Local Broadcast Journalism. *Limitation: A short-form news dispatch relying heavily on police department press statements.*  
21  
Illinois State Police. "Overview \- Tamara Clayton Expressway Camera Act." Illinois State Police, 2024\. URL: https://isp.illinois.gov/CriminalInvestigations/TransparencyPage. Accessed: July 24, 2026\. Source type: Official Government Portal. *Limitation: Provides legislative background strictly from the perspective of the implementing state agency.*  
30  
House Democratic Cloakroom. "Wednesday, April 17, 2024, Daily Floor Review." U.S. House of Representatives, April 17, 2024\. URL: https://democraticcloakroom.house.gov/about/events/wednesday-april-17-2024-daily-floor-review. Accessed: July 24, 2026\. Source type: Government Legislative Tracker. *Limitation: A raw parliamentary log of bill passage without analytical text or context.*  
31  
Congressional Research Service. "Fourth Amendment is Not For Sale Act." CRS Report R48592, 2024\. URL: https://www.everycrsreport.com/reports/R48592.html. Accessed: July 24, 2026\. Source type: Government Research Report. *Limitation: Summarizes legislative intent; legislation ultimately failed to pass the Senate.*  
6  
Rocher, L., Hendrickx, J.M., de Montjoye, Y.-A. "Estimating the success of re-identifications in incomplete datasets using generative models." Nature Communications 10, 3069, July 23, 2019\. URL: https://pubmed.ncbi.nlm.nih.gov/31337762/. Accessed: July 24, 2026\. Source type: Peer-Reviewed Scientific Journal. *Limitation: Relies on probabilistic modeling rather than real-world adversarial attacks.*  
7  
Rocher, L., Hendrickx, J.M., de Montjoye, Y.-A. "Estimating the success of re-identifications in incomplete datasets using generative models." Nature Communications, 2019\. URL: https://www.researchgate.net/publication/334634583. Accessed: July 24, 2026\. Source type: Peer-Reviewed Scientific Journal. *Limitation: Identical to source 63; represents the open-access manuscript version.*  
8  
ScienceDaily / Imperial College London. "Anonymizing personal data 'not enough to protect privacy,' shows new study." ScienceDaily, July 23, 2019\. URL: https://www.sciencedaily.com/releases/2019/07/190723110523.htm. Accessed: July 24, 2026\. Source type: Science Journalism / Press Release. *Limitation: A simplified summary of complex statistical methodologies intended for a lay audience.*  
9  
Imperial College London / Written Evidence to UK Parliament. "Written Evidence 43358." UK Parliament Committees, January 2022\. URL: https://committees.parliament.uk/writtenevidence/43358/html/. Accessed: July 24, 2026\. Source type: Parliamentary Evidence Submission. *Limitation: Represents policy advocacy from researchers based on their technical findings.*  
10  
de Montjoye, Y.-A., Hidalgo, C.A., Verleysen, M., Blondel, V.D. "Unique in the Crowd: The privacy bounds of human mobility." Scientific Reports 3, 1376, March 2013\. URL: https://www.researchgate.net/publication/236076438. Accessed: July 24, 2026\. Source type: Peer-Reviewed Scientific Journal. *Limitation: Focuses exclusively on spatial-temporal mobility data, predating modern generative models.*  
11  
de Montjoye, Y.-A., et al. "Unique in the Crowd: The Privacy Bounds of Human Mobility." Scientific Reports 3 (1): 1–5, 2013\. URL: https://www.tandfonline.com/doi/full/10.1080/17538947.2021.1952324. Accessed: July 24, 2026\. Source type: Peer-Reviewed Scientific Journal. *Limitation: Foundational research that has since been expanded upon by 2019 copula-based studies.*

#### **Works cited**

> 1. Data brokers: A call for transparency and accountability \- ResearchGate, [https://www.researchgate.net/publication/292937377\_Data\_brokers\_A\_call\_for\_transparency\_and\_accountability](https://www.researchgate.net/publication/292937377_Data_brokers_A_call_for_transparency_and_accountability)  
> 2. Data Brokers: A Call for Transparency and Accountability Opening Remarks of Chairwoman Edith Ramirez, Washington, DC \- Federal Trade Commission, [https://www.ftc.gov/system/files/documents/public\_statements/311891/140527databrokers.pdf](https://www.ftc.gov/system/files/documents/public_statements/311891/140527databrokers.pdf)  
> 3. FTC Data Broker Report Calls for More Transparency and Consumer Control | Inside Privacy, [https://www.insideprivacy.com/uncategorized/ftc-data-broker-report-calls-for-more-transparency-and-consumer-control/](https://www.insideprivacy.com/uncategorized/ftc-data-broker-report-calls-for-more-transparency-and-consumer-control/)  
> 4. Data Brokers: A Look at the Canadian and American Landscape \- September 2014 \- Office of the Privacy Commissioner of Canada, [https://www.priv.gc.ca/en/opc-actions-and-decisions/research/explore-privacy-research/2014/db\_201409/?wbdisable=true](https://www.priv.gc.ca/en/opc-actions-and-decisions/research/explore-privacy-research/2014/db_201409/?wbdisable=true)  
> 5. Data Brokers: A Call for Transparency and Accountability Matter No. P125404 Statement of Commissioner Julie Brill \- Federal Trade Commission, [https://www.ftc.gov/system/files/documents/public\_statements/311551/140527databrokerrptbrillstmt.pdf](https://www.ftc.gov/system/files/documents/public_statements/311551/140527databrokerrptbrillstmt.pdf)  
> 6. Estimating the success of re-identifications in incomplete datasets using generative models \- PubMed, [https://pubmed.ncbi.nlm.nih.gov/31337762/](https://pubmed.ncbi.nlm.nih.gov/31337762/)  
> 7. (PDF) Estimating the success of re-identifications in incomplete datasets using generative models \- ResearchGate, [https://www.researchgate.net/publication/334634583\_Estimating\_the\_success\_of\_re-identifications\_in\_incomplete\_datasets\_using\_generative\_models](https://www.researchgate.net/publication/334634583_Estimating_the_success_of_re-identifications_in_incomplete_datasets_using_generative_models)  
> 8. Anonymizing personal data 'not enough to protect privacy,' shows new study | ScienceDaily, [https://www.sciencedaily.com/releases/2019/07/190723110523.htm](https://www.sciencedaily.com/releases/2019/07/190723110523.htm)  
> 9. We'd like to use additional cookies to understand how you use the site and improve our services. \- UK Parliament Committees, [https://committees.parliament.uk/writtenevidence/43358/html/](https://committees.parliament.uk/writtenevidence/43358/html/)  
> 10. (PDF) Unique in the Crowd: The Privacy Bounds of Human Mobility \- ResearchGate, [https://www.researchgate.net/publication/236076438\_Unique\_in\_the\_Crowd\_The\_Privacy\_Bounds\_of\_Human\_Mobility](https://www.researchgate.net/publication/236076438_Unique_in_the_Crowd_The_Privacy_Bounds_of_Human_Mobility)  
> 11. Full article: Human mobility data in the COVID-19 pandemic: characteristics, applications, and challenges \- Taylor & Francis, [https://www.tandfonline.com/doi/full/10.1080/17538947.2021.1952324](https://www.tandfonline.com/doi/full/10.1080/17538947.2021.1952324)  
> 12. Case: Louis v. SafeRent Solutions, LLC \- Civil Rights Litigation Clearinghouse, [https://clearinghouse.net/case/45888/](https://clearinghouse.net/case/45888/)  
> 13. Memorandum and Order \- Mary Louis v. Saferent Solutions, LLC (D. Mass.) \- Department of Justice, [https://www.justice.gov/crt/media/1310736/dl](https://www.justice.gov/crt/media/1310736/dl)  
> 14. Landlord Tenant Screening \- SmartScreen, [https://smartscreen.clearscreening.com/landlord-tenant-screening/](https://smartscreen.clearscreening.com/landlord-tenant-screening/)  
> 15. Louis et al. v. SafeRent et al. (D. Mass.) | United States Department of Justice, [https://www.justice.gov/crt/case/louis-et-al-v-saferent-et-al-d-mass](https://www.justice.gov/crt/case/louis-et-al-v-saferent-et-al-d-mass)  
> 16. Civil Rights Advocates Achieve the Nation's Strongest Police Department Policy on Facial Recognition Technology \- ACLU of Michigan, [https://www.aclumich.org/press-releases/civil-rights-advocates-achieve-nations-strongest-police-department-policy-facial/](https://www.aclumich.org/press-releases/civil-rights-advocates-achieve-nations-strongest-police-department-policy-facial/)  
> 17. Facial recognition false arrest of man by Detroit police wins settlement, [https://www.fox2detroit.com/news/facial-recognition-false-arrest-man-detroit-police-wins-settlement](https://www.fox2detroit.com/news/facial-recognition-false-arrest-man-detroit-police-wins-settlement)  
> 18. [https://clearinghouse.net/case/44074/\#:\~:text=On%20July%2012%2C%202024%2C%20the,the%20ADEA%2C%20and%20Section%201981.](https://clearinghouse.net/case/44074/#:~:text=On%20July%2012%2C%202024%2C%20the,the%20ADEA%2C%20and%20Section%201981.)  
> 19. Mobley v. Workday, Inc. | 3:2023cv00770 | N.D. Cal ... \- CaseMine, [https://www.casemine.com/judgement/us/6a4c9adae971d2e88f2d9983](https://www.casemine.com/judgement/us/6a4c9adae971d2e88f2d9983)  
> 20. Cicero Police Department opens new real-time crime center \- CBS Chicago, [https://www.cbsnews.com/chicago/news/cicero-police-department-new-real-time-crime-center/](https://www.cbsnews.com/chicago/news/cicero-police-department-new-real-time-crime-center/)  
> 21. Automated License Plate Trader \- Transparency Page \- Division of Criminal Investigation \- Illinois State Police, [https://isp.illinois.gov/CriminalInvestigations/TransparencyPage](https://isp.illinois.gov/CriminalInvestigations/TransparencyPage)  
> 22. FTC v Kochava, Inc. | Federal Trade Commission, [https://data.aclum.org/storage/2025/01/FTC\_www\_ftc\_gov\_legal-library\_browse\_cases-proceedings\_ftc-v-kochava-inc.pdf](https://data.aclum.org/storage/2025/01/FTC_www_ftc_gov_legal-library_browse_cases-proceedings_ftc-v-kochava-inc.pdf)  
> 23. FTC Commercial Data Surveillance Crack Down: Kochava Gains a Win in Data Privacy Suit, [https://www.akingump.com/en/insights/blogs/ag-data-dive/ftc-commercial-data-surveillance-crack-down-kochava-gains-a-win-in-data-privacy-suit](https://www.akingump.com/en/insights/blogs/ag-data-dive/ftc-commercial-data-surveillance-crack-down-kochava-gains-a-win-in-data-privacy-suit)  
> 24. Gravy Analytics, Inc.; Analysis of Proposed Consent Order to Aid Public Comment, [https://www.federalregister.gov/documents/2024/12/06/2024-28738/gravy-analytics-inc-analysis-of-proposed-consent-order-to-aid-public-comment](https://www.federalregister.gov/documents/2024/12/06/2024-28738/gravy-analytics-inc-analysis-of-proposed-consent-order-to-aid-public-comment)  
> 25. Misinterpreting Section 5(n) of the FTC Act: A Critique of the District Court's Rulings in FTC v. Kochava | Journal of Law and Commerce \- University of Pittsburgh, [https://jlc.law.pitt.edu/ojs/jlc/article/view/296](https://jlc.law.pitt.edu/ojs/jlc/article/view/296)  
> 26. Federal Trade Commission v. Kochava, Inc., No. 2:2022cv00377 \- Document 101 (D. Idaho 2025), [https://law.justia.com/cases/federal/district-courts/idaho/iddce/2:2022cv00377/50683/101](https://law.justia.com/cases/federal/district-courts/idaho/iddce/2:2022cv00377/50683/101)  
> 27. FTC's Kochava Settlement Advances Data Privacy Enforcement–But Leaves Critical Gaps in Protecting Consumers \- Yale Journal on Regulation, [https://www.yalejreg.com/nc/ftcs-kochava-settlement-advances-data-privacy-enforcement-but-leaves-critical-gaps-in-protecting-consumers/](https://www.yalejreg.com/nc/ftcs-kochava-settlement-advances-data-privacy-enforcement-but-leaves-critical-gaps-in-protecting-consumers/)  
> 28. A view from DC: Kochava is not enough \- IAPP, [https://iapp.org/news/a/a-view-from-dc-kochava-is-not-enough](https://iapp.org/news/a/a-view-from-dc-kochava-is-not-enough)  
> 29. The FTC's inadequate Kochava proposed privacy settlement \- CLP Blog, [https://clpblog.citizen.org/the-ftcs-inadequate-kochava-proposed-privacy-settlement/](https://clpblog.citizen.org/the-ftcs-inadequate-kochava-proposed-privacy-settlement/)  
> 30. Wednesday, April 17, 2024, Daily Floor Review \- Democratic Cloakroom \- House.gov, [https://democraticcloakroom.house.gov/about/events/wednesday-april-17-2024-daily-floor-review](https://democraticcloakroom.house.gov/about/events/wednesday-april-17-2024-daily-floor-review)  
> 31. FISA Section 702 and the 2024 Reforming Intelligence and Securing America Act \- EveryCRSReport.com, [https://www.everycrsreport.com/reports/R48592.html](https://www.everycrsreport.com/reports/R48592.html)  
> 32. Disparate Impact as Uniquely Relevant in the Age of AI, [https://civilrights.org/disparate-impact-age-of-ai/](https://civilrights.org/disparate-impact-age-of-ai/)  
> 33. EU AI Act (Regulation 2024/1689) | Comparative AI, [https://comparativeai.org/rules/eu/ai-act/](https://comparativeai.org/rules/eu/ai-act/)  
> 34. Data Brokers: A Call For Transparency and Accountability: A Report of the Federal Trade Commission (May 2014), [https://www.ftc.gov/system/files/documents/reports/data-brokers-call-transparency-accountability-report-federal-trade-commission-may-2014/140527databrokerreport.pdf](https://www.ftc.gov/system/files/documents/reports/data-brokers-call-transparency-accountability-report-federal-trade-commission-may-2014/140527databrokerreport.pdf)