Interference from Behavioral Advertising and Behavioral Profiling
5 minutes
5 Questions
In the CIPT body of knowledge, interference refers to privacy harms where an organization's actions disrupt an individual's life, solitude or autonomy. Drawing on Daniel Solove's taxonomy, interference includes intrusion, meaning invasive acts that disturb a person's tranquility or solitude, and de…In the CIPT body of knowledge, interference refers to privacy harms where an organization's actions disrupt an individual's life, solitude or autonomy. Drawing on Daniel Solove's taxonomy, interference includes intrusion, meaning invasive acts that disturb a person's tranquility or solitude, and decisional interference, meaning outside influence over personal decisions. Behavioral advertising and behavioral profiling are major technological sources of interference.
Behavioral advertising tracks users' online activities, such as browsing history, search queries, clicks, app usage, location and purchases. It relies on tools like cookies, tracking pixels, device fingerprinting, SDKs and cross-device identifiers, and uses the resulting data to deliver targeted ads. Interference arises when ads follow users across sites (retargeting), appear at intrusive moments, or reveal sensitive inferences, such as a pregnancy or health condition, to others sharing a device. Users often feel watched, which chills behavior and erodes trust.
Behavioral profiling goes further. It aggregates data from multiple sources to build detailed models that predict interests, traits, vulnerabilities and future behavior. Profiles can enable manipulation through personalized pricing, dark patterns or emotionally exploitative content. This is decisional interference because it undermines autonomy. Profiling can also cause discrimination, such as exclusion from credit, housing or employment opportunities. It may also produce inaccurate inferences that individuals cannot see or correct.
From a privacy risk management perspective, technologists should identify these risks through privacy impact assessments and threat modeling. Appropriate controls include:
- obtaining meaningful consent, with opt-in for sensitive data
- honoring opt-outs and browser signals such as Global Privacy Control
- minimizing data collection and retention
- avoiding targeting based on sensitive categories
- preferring contextual advertising over behavioral advertising where feasible
- applying aggregation or differential privacy
- providing clear notices and preference dashboards
Several frameworks shape these obligations. Industry self-regulatory programs (DAA, NAI) set baseline practices. The GDPR, including Article 22 on automated decision-making, and the ePrivacy Directive's cookie consent rules apply in Europe. The CCPA/CPRA regulates sharing data for cross-context behavioral advertising.
Ultimately, the goal is to respect user expectations and autonomy while still delivering legitimate business value.
Interference from Behavioral Advertising and Behavioral Profiling: A CIPT Guide to Privacy Risk Management
Overview In the CIPT Body of Knowledge, interference is a category of privacy harm in which an organization's actions intrude on a person's life or disrupt their ability to make free and autonomous decisions. Behavioral advertising and behavioral profiling are the textbook examples. They show how technology that seems harmless, such as cookies, analytics and recommendation engines, can become a privacy threat. This guide covers why the topic matters, what it is, how it works technically, which controls reduce the risk, and how to answer exam questions about it.
1. Why It Is Important
It is a core CIPT risk concept. The CIPT exam tests whether technologists can recognize privacy harms that go beyond data breaches. Interference is a harm that can occur even when no data is leaked or stolen.
It is pervasive. Almost every website, mobile app, smart TV, connected car and IoT device feeds data into advertising and profiling ecosystems. Technologists build and integrate these systems every day, often through third-party SDKs and tags.
It erodes autonomy. Profiling can shape what people see, what prices they are offered, which opportunities they learn about, and how they are judged. This is called decisional interference.
It creates regulatory exposure. Relevant laws and frameworks include:
GDPR: Articles 21 and 22 on objection to direct marketing and on automated decision-making and profiling.
ePrivacy Directive: cookie consent requirements.
CCPA/CPRA: the right to opt out of the sale or sharing of personal information for cross-context behavioral advertising, plus honoring the Global Privacy Control signal.
US state privacy laws: opt-outs for targeted advertising and profiling.
COPPA: rules on children's data.
FTC Act Section 5: enforcement against deceptive or unfair practices.
It damages trust. Well-known incidents show the reputational cost of profiling that feels creepy:
Target: its pregnancy-prediction model reportedly revealed a teenager's pregnancy to her family.
Facebook Beacon: it broadcast users' off-site purchases to their friends.
Cambridge Analytica: it used psychographic profiling for political targeting.
2. What It Is Behavioral advertising (also called online behavioral advertising, OBA, interest-based or targeted advertising) is tracking an individual's activity over time, often across unrelated websites, apps and devices, in order to deliver ads matched to inferred interests.
Behavioral profiling is the broader practice of collecting and analyzing data about a person's behavior to infer characteristics, preferences, or likely future actions. Examples of inferences include health status, sexual orientation, financial distress, political views, pregnancy, or propensity to churn. The profile is then used to make decisions about the person or to influence them.
Where it fits in privacy risk frameworks:
Solove's Taxonomy of Privacy groups harms into four categories: information collection, information processing, information dissemination and invasion. Interference maps mainly to invasion, which has two forms:
Intrusion: unwanted incursions into a person's life or solitude. Examples are intrusive retargeting ads that follow a user everywhere, or ads that reveal sensitive inferences on a shared screen.
Decisional interference: government or organizational influence over a person's private decisions. Examples are manipulative targeting, price discrimination and exploiting vulnerabilities.
Profiling also involves processing harms from the taxonomy:
Aggregation: combining scattered data points into a revealing profile.
Secondary use: using data collected for one purpose for an unrelated one.
Exclusion: individuals not knowing about the profile and having no say in it.
Increased accessibility and disclosure: when profiles are shared or exposed.
Calo's harm dimensions:
Subjective harm is the perception of being watched, the creepiness effect.
Objective harm is the actual adverse use of information, such as being denied credit or charged more.
Contextual integrity (Nissenbaum): tracking a user's health-site visits and using that data for insurance or retail ads violates the norms of the context in which the data was shared.
3. How It Works a) Data collection and tracking technologies
First-party cookies are set by the site the user visits. Third-party cookies are set by ad networks or trackers embedded in the page, which allows tracking across sites.
Web beacons (pixels, clear GIFs) are tiny embedded images or scripts that report page views, email opens and conversions.
Local storage and similar mechanisms: HTML5 local storage, Flash LSOs (historically), ETags and so-called zombie or evercookies that respawn after deletion.
Device and browser fingerprinting identifies a device from its attributes, such as fonts, screen size, user agent and canvas rendering, with no stored identifier at all.
Mobile advertising IDs such as Apple's IDFA and Google's AAID, plus SDKs embedded in apps.
Cross-device tracking:
Deterministic matching links devices through logins.
Probabilistic matching links devices through shared IP addresses, location and usage patterns.
Offline and data broker enrichment: purchase history, loyalty cards, public records, location data and hashed email onboarding.
b) The ad-tech ecosystem
Publishers, advertisers, ad networks, ad exchanges, supply-side platforms (SSPs), demand-side platforms (DSPs) and data management platforms (DMPs) all take part.
Real-time bidding (RTB) broadcasts bid requests containing user data, such as location, device and interest segments, to many potential bidders in milliseconds. This raises concerns about uncontrolled dissemination.
Cookie syncing lets different trackers match their identifiers for the same user.
c) Profiling and inference
Raw behavior, such as clicks, searches, purchases, location and dwell time, is turned into segments. Examples are 'new parent', 'auto intender' or 'diabetes interest'.
Machine learning models predict attributes the user never disclosed. These inferred data can be more sensitive than the inputs, which is the core aggregation risk.
d) Resulting interference harms
Manipulation: targeting people at moments of vulnerability, for example gambling ads to problem gamblers or payday loans to people in financial distress, and using dark patterns.
Discrimination and exclusion: housing, job or credit ads shown or hidden by protected characteristics, and price discrimination.
Filter bubbles: narrowing the information people see, including political microtargeting.
Unwanted disclosure: sensitive ads appearing on shared devices, which reveals inferences to others.
Chilling effects: people changing their behavior because they feel watched.
Loss of control: opaque data flows to parties the user never knowingly dealt with.
4. Controls and Mitigations Transparency and notice
Clear privacy notices and layered or just-in-time notices.
The AdChoices icon from the DAA self-regulatory program, plus NAI and EDAA principles.
Explanations such as 'Why am I seeing this ad?'.
Choice and consent
Opt-in consent for non-essential cookies under the ePrivacy Directive and GDPR, managed through a consent management platform (CMP) and frameworks such as the IAB TCF.
Opt-out mechanisms, including Do Not Sell or Share links.
Honoring browser signals: Global Privacy Control, which is legally required in some US states, and Do Not Track, which is historically voluntary and largely ignored.
Express consent for sensitive categories such as health, children's data and precise location.
Data minimization and purpose limitation
Collect only what is needed.
Use contextual advertising, which targets based on page content rather than the user's history.
Set short retention periods.
Do not repurpose data without a lawful basis.
Sensitive segment restrictions
Prohibit targeting on health, sexual orientation, children, race, religion and similar characteristics.
Prevent discriminatory delivery for housing, employment and credit ads.
De-identification and privacy-enhancing technologies (PETs)
Aggregation, k-anonymity and differential privacy.
On-device processing and federated learning.
Privacy-preserving measurement and data clean rooms.
Be careful: hashed identifiers are still personal data, because they are pseudonymous, not anonymous.
Rights under automated decision-making rules (GDPR Article 22)
The right not to be subject to solely automated decisions with legal or similarly significant effects.
Human review and meaningful information about the logic involved.
Governance
Data Protection Impact Assessments (DPIAs). Large-scale profiling is typically considered high risk.
Vendor and SDK due diligence, data processing agreements and regular tag audits.
Privacy by Design and by Default.
Platform-level changes
Browser restrictions on third-party cookies, such as Safari ITP and Firefox ETP.
Apple's App Tracking Transparency (ATT) prompt.
5. How to Answer Exam Questions on This Topic CIPT questions are usually scenario-based. A typical question describes a company deploying tracking, recommendation or targeting technology and asks you to:
identify the type of privacy harm,
choose the best technical or process control, or
spot the legal or ethical problem.
Step-by-step approach:
Classify the harm. Ask what is happening to the individual.
Being followed or bothered points to intrusion.
Being influenced, steered, excluded or manipulated in personal choices points to decisional interference.
Many data points combined into a revealing picture points to aggregation.
Data used beyond its original purpose points to secondary use.
Identify the technology. Is it third-party cookies, fingerprinting, mobile IDs, cross-device tracking or RTB? Remember that fingerprinting evades cookie deletion, so cookie-based controls alone do not solve it.
Check sensitivity and audience. Health, children, precise location, financial status and protected classes raise the risk sharply and usually call for opt-in consent or prohibition.
Map to the legal requirement.
EU: prior consent for non-essential cookies, a lawful basis, Article 22 for significant automated decisions, and a DPIA.
US: notice, opt-out of sale, sharing or targeted ads, GPC, COPPA for children under 13, and FTC rules on deception and unfairness.
Pick the most effective, privacy-protective control. Prefer to prevent risk at the source rather than merely disclose it. The general order of preference is:
Avoid or minimize collection, for example by using contextual ads.
Obtain meaningful consent.
Apply de-identification or PETs.
Provide notice and opt-out.
Exam Tips: Answering Questions on Interference from Behavioral Advertising and Behavioral Profiling
Memorize the vocabulary. Know intrusion and decisional interference, the two invasion harms in Solove's taxonomy. If a question asks which harm occurs when targeted ads steer a vulnerable person's choices, the answer is decisional interference.
Interference does not require a breach. If no data was leaked but the user was manipulated, embarrassed or excluded, the harm is interference, not a security failure.
Inferences are personal data. Derived attributes, such as predicted pregnancy or inferred health conditions, carry the same or greater sensitivity as the source data. Watch for answers that wrongly treat inferences as harmless.
Hashed is not anonymous. Hashed emails and advertising IDs are pseudonymous identifiers that still enable tracking. Reject any answer that calls them anonymized.
Notice alone is rarely the best answer. When the options include a long privacy policy versus minimizing collection or getting consent, the stronger control usually wins.
Know which controls are weak.
Do Not Track was voluntary and widely ignored.
Deleting cookies does not stop fingerprinting.
Industry self-regulatory opt-outs (DAA, NAI) often rely on opt-out cookies, which are lost when cookies are cleared.
Know which controls are strong. Contextual advertising, on-device processing, consent before tracking, sensitive-category bans, short retention, honoring GPC, DPIAs and vendor and tag audits.
Third parties are your responsibility. Embedding an ad SDK or tracking pixel makes the organization accountable for the data flows it creates. The best answer often includes vendor due diligence, contracts and limiting what is shared.
Children trigger special rules. Behavioral advertising to children under 13 in the US requires verifiable parental consent under COPPA. The GDPR also gives children's data heightened protection.
Spot significant effects. When profiling affects credit, employment, housing, insurance or education, think GDPR Article 22, human review, explainability and anti-discrimination concerns.
Use the creepiness test. Questions describing users feeling surprised or uncomfortable are pointing to a violation of contextual integrity or reasonable expectations. Choose controls that realign the practice with user expectations through transparency, choice and limiting use.
Read for the BEST answer. Several options may be technically correct. Prefer the one that addresses the root cause, follows Privacy by Design, and is proactive rather than reactive.
Remember the classic cases. Target pregnancy prediction (inference and aggregation), Facebook Beacon (disclosure without consent), Cambridge Analytica (secondary use and political manipulation) and discriminatory ad delivery (exclusion and decisional interference) are frequently used as analogies.
Quick Summary Behavioral advertising and profiling track individuals across contexts and infer sensitive traits. They then use those inferences to intrude on people or to influence their decisions. These are interference harms in Solove's taxonomy, specifically intrusion and decisional interference, and they are compounded by aggregation and secondary use. Mitigate them with minimization, consent, sensitive-data restrictions, PETs, transparency, opt-out signals, DPIAs and vendor oversight. On the exam, classify the harm, identify the technology, weigh sensitivity, map to the law, and choose the most preventive control.
