Skip to main content
Be Bitwise

Data Brokers & Regulatory Landscape

Mobile Privacy & Data Protection · Mobile Advanced Security track

12 min read

In this module

The Data Supply Chain

Phones produce a relentless trickle of personal data — location, behaviour, interests, contacts, device characteristics. None of it stays with the app that scraped it. Instead it flows through a sprawling supply chain: ad networks, data brokers, aggregators, analytics firms. Each link tags on a markup and resells.

The Mobile Data Pipeline

From the user's tap to the buyer's database, the path is depressingly predictable:

  1. Collection: Mobile apps embed advertising and analytics SDKs that collect device data (location, advertising ID, app usage, device attributes).
  2. Aggregation: SDK providers aggregate data from thousands of apps, building rich profiles linked to advertising IDs or device fingerprints.
  3. Brokerage: Data aggregators sell or licence their datasets to data brokers, who combine data from multiple sources, enrich it with public records and purchase data, and repackage it for sale.
  4. Consumption: End buyers include advertisers (for targeted campaigns), hedge funds (for market intelligence), insurance companies (for risk assessment), and government agencies (for surveillance and intelligence).

Real-Time Bidding and Location Leakage

Real-time bidding (RTB) is the engine of programmatic advertising. It's also one of the world's biggest accidental location-data firehoses.

How RTB Works

A mobile app is about to show an ad. Here's what happens behind the scenes:

  1. The ad SDK fires a bid request at the ad exchange.
  2. That request packs a lot in: GPS coordinates (often six decimal places), advertising ID, device model, OS version, carrier, app name, content context.
  3. The exchange fans the request out to hundreds of demand-side platforms (DSPs) and ad networks.
  4. Each DSP looks at the request and decides whether to bid.
  5. Highest bidder wins and serves the ad.

The Data Leakage Problem

Here's the kicker: every participant in the auction sees the full bid request, GPS coordinates and device identifier included. Most of them lose the auction. They still keep the location data.

So any company that joins RTB auctions has a live feed of location data for millions of devices. A few companies participate purely to harvest the data, with no genuine interest in placing ads.

Follow one ad slot through an auction. Three DSPs stand in for the hundreds a real exchange contacts, and the field names come from the OpenRTB spec most exchanges use. Count how many copies of the location exist by the end, and how many of them bought anything.

RTB as a Surveillance Infrastructure

The RTB ecosystem inadvertently created one of the world's largest surveillance infrastructures. Bid requests containing precise GPS coordinates and device identifiers are broadcast to hundreds of companies billions of times per day. Any participant can collect and store this data. Investigative reporting has revealed that data brokers, intelligence contractors, and government agencies have exploited RTB data streams to conduct mass surveillance without obtaining warrants or court orders, because the data is classified as "commercially available information".

Case Studies: Government Purchase of Location Data

Locate X

Locate X is a product from Babel Street, a data analytics outfit. It hands you a visual interface for querying a database of mobile location data. Draw a geofence on a map and pull back the device identifiers and movement histories of every device that passed through during the chosen window.

The data underneath comes from the commercial advertising ecosystem — the same bid-request stream described above. According to procurement records and press reporting (Motherboard, later 404 Media), multiple US agencies (ICE, CBP, Secret Service, IRS) have bought licences to Locate X and used it to track individuals without applying for a location warrant.

Babel Street

Babel Street and its peers (Venntel, Anomaly Six, X-Mode/Outlogic) live in a legal grey zone. Their argument: the location data was collected with user "consent" (buried somewhere on page 14 of an app privacy policy) and is on sale on the open market, so government use of it shouldn't need a warrant. That reading is contested, and lawsuits are working through the courts. Regulators have started to move as well. Since 2024 the US Federal Trade Commission has reached orders with X-Mode/Outlogic and InMarket, and proposed orders against Gravy Analytics (Venntel's parent) and Mobilewalla, restricting the sale of sensitive location data such as visits to clinics and places of worship. It also sued Kochava over the same kind of data in 2022.

The Carpenter Decision

In Carpenter v. United States (2018), the US Supreme Court ruled that pulling historical cell-site location information counts as a search under the Fourth Amendment and needs a warrant. The government's response: commercially purchased location data (from brokers, not carriers) sits outside Carpenter because the user "voluntarily" shared it with commercial entities. The distinction between carrier-held data and commercially purchased data that reveals exactly the same movements remains legally unresolved. Carpenter built on Riley v. California (2014), where the Court held that police need a warrant to search the contents of a phone seized during an arrest.

GDPR and Mobile

The General Data Protection Regulation (GDPR) has been in force since May 2018 and casts the widest net of any data-protection law touching mobile apps.

Key Requirements for Mobile

  • Lawful basis: Any processing of personal data requires a lawful basis. For advertising tracking, consent is typically the required basis - and that consent must be freely given, specific, informed, and unambiguous.
  • Consent requirements: Pre-ticked boxes, bundled consent ("accept all to use the app"), and consent walls (blocking access unless consent is granted) are prohibited. Each SDK's data collection requires its own specific consent.
  • Data subject rights: Users have the right to access all data held about them, correct inaccuracies, request erasure ("right to be forgotten"), data portability (receive their data in a machine-readable format), and object to processing.
  • Data Protection Impact Assessments: Required for high-risk processing such as large-scale location tracking or systematic monitoring.
  • Fines: Up to 4% of global annual revenue or 20 million euros, whichever is higher.

Mobile-Specific GDPR Challenges

  • SDK consent: An app with ten advertising SDKs needs consent for each SDK's data collection. Most consent management platforms present this as a single dialog, but GDPR requires granular, per-purpose consent.
  • Advertising ID as personal data: The Article 29 Working Party (now EDPB) confirmed that advertising identifiers constitute personal data under GDPR because they can identify an individual's device.
  • Cross-border data flows: Mobile data collected in the EU and transferred to US-based ad networks raises data transfer compliance issues.
  • The consent machinery itself: In 2022 the Belgian data protection authority found that IAB Europe's Transparency and Consent Framework, the system behind most ad-tech consent pop-ups and the consent signals passed through real-time bidding, breached GDPR, and fined IAB Europe €250,000. In 2024 the EU Court of Justice backed the key points: the consent string can be personal data, and IAB Europe can be a joint controller.

CCPA/CPRA

The California Consumer Privacy Act (CCPA), expanded by the California Privacy Rights Act (CPRA), hands California residents a set of rights over their mobile data.

Key Provisions

  • Right to know: Consumers can request disclosure of what personal information is collected, sold, or shared.
  • Right to delete: Consumers can request deletion of personal information.
  • Right to opt-out of sale or sharing: Consumers can direct businesses not to sell their personal information, or share it for cross-context behavioural advertising. Businesses must display a "Do Not Sell or Share My Personal Information" link.
  • Sensitive personal information: CPRA added a category for "sensitive personal information" that includes precise geolocation. Consumers get a right to limit its use to what the service needs, exercised through a "Limit the Use of My Sensitive Personal Information" link. That's an opt-out, not opt-in consent; opt-in applies to selling the data of consumers under 16.

Children's Privacy: COPPA

The Children's Online Privacy Protection Act (COPPA) clamps down hard on apps and services aimed at children under 13:

  • Verifiable parental consent is required before collecting personal information from children.
  • No behavioural advertising to children.
  • Minimisation: Collect only what is necessary for the service.
  • App store age-gating: Both Apple and Google require developers to declare whether their apps are directed at children and to comply with COPPA and equivalent laws.

The penalties bite. Google handed over $170 million to settle FTC allegations that YouTube was collecting children's data without parental consent. The FTC also tightened the COPPA Rule in 2025: among other changes, sharing children's data with third parties for targeted advertising now needs its own separate parental consent.

Try placing three apps against the rules above. For each, which regimes apply, and what's one control it must have?

  1. A free weather app, based in Ireland, that sells precise location to an ad exchange. Users worldwide, including California.
  2. A cartoon puzzle game aimed at 8-year-olds in the US, with an ad SDK.
  3. A US period-tracking app with EU users that stores cycle data on its servers.
Show the answer
  1. GDPR (EU establishment) and CCPA/CPRA for its California users. Under GDPR it needs specific, freely given consent before the location leaves the phone; for California it needs the "Do Not Sell or Share" link, and precise location is sensitive personal information, so the "Limit the Use" right applies too.
  2. COPPA. Verifiable parental consent before collecting personal information, and the ad SDK has to run in its child-directed mode with behavioural advertising off. The store listing must declare it's aimed at children.
  3. GDPR, where health data is special-category, so it needs explicit consent and a DPIA. If it has California users, CPRA treats health data as sensitive too. State health-data laws such as Washington's My Health My Data Act may also reach it.

Privacy by Design

Privacy by design wears two hats: regulatory requirement (per GDPR) and engineering principle. For mobile apps, it lands as a set of concrete technical choices.

Data Minimisation

Collect only what the service genuinely needs. A weather app wants approximate location, city-level. It does not need GPS to six decimals. A fitness app wants step counts, not 24-hour background location tracking.

On-Device Processing

Crunch the data on the phone instead of shipping it to a server. Apple's Core ML and Create ML frameworks make on-device inference straightforward. Federated learning lets you improve models without ever centralising raw data.

Differential Privacy

Sprinkle calibrated noise into the data so individual contributions vanish into the static. Apple does this with iOS telemetry: the aggregated stats are useful, but no single device's traces survive.

The oldest version of the idea is randomised response, from 1960s survey research. Each phone flips a coin before answering: heads, it tells the truth; tails, it flips again and reports that instead. Any single yes might be a coin, yet across thousands of reports the server can subtract the noise back out:

Try itDrop users to 100 and watch the estimate wobble. Then make the first coin tell the truth three times in four and see accuracy rise as deniability falls.
C (GCC 9.2.0)

Ephemeral Processing

Process in memory, then drop the data. If you don't need to keep it for the service to work, don't keep it.

iOS Privacy Nutrition Labels and Google Data Safety

Both Apple and Google now make developers disclose their data collection practices in a standardised format that ends up on the app store listing.

Apple Privacy Nutrition Labels

Introduced in December 2020, Apple's labels categorise data into types (location, contacts, identifiers, usage data) and purposes (tracking, analytics, advertising, app functionality). Each type is marked as "linked to you" or "not linked to you". Labels are displayed before download.

Google Data Safety

Google's Data Safety section (launched 2022) requires similar disclosures on the Play Store. Developers declare what data is collected, whether it is shared with third parties, and what security practices are in place (encryption, deletion on request).

Limitations

Both systems rely on developer self-reporting, which is the obvious failure mode. Researchers have repeatedly caught apps collecting data they never declared on the label, just by watching the network traffic. Neither Apple nor Google audits these declarations systematically. Both reserve the right to pull apps caught lying, but the reservation only matters if the inaccuracies surface.

Key Concept
Mobile Privacy Regulatory Landscape

Mobile privacy is governed by a patchwork of regulations: GDPR (EU - consent, data subject rights, 4% revenue fines), CCPA/CPRA (California - right to know, delete, opt-out of sale), COPPA (US - children's privacy, parental consent). Privacy by design principles (data minimisation, on-device processing, differential privacy) provide technical implementation guidance. App store privacy labels (Apple Privacy Nutrition Labels, Google Data Safety) provide transparency but rely on self-reporting. The gap between regulatory requirements and actual industry practice remains significant, particularly in the data broker and RTB ecosystems where location data flows through complex supply chains with limited user visibility or control.