Not every tracking signal exists to sell you shoes.
Some exist because somebody is trying to steal the shoe advertiser’s money.
Digital advertising has a genuine fraud problem: bots can generate fake impressions, automated systems can click ads, fraudulent publishers can misrepresent inventory, and traffic can be manipulated to look human long enough to get paid.
Detecting that abuse requires observation.
The privacy question is what happens after the observation is collected.
Fraud detection needs signals
A fraud system may examine patterns such as:
- unusually rapid clicks,
- impossible or suspicious traffic volumes,
- repeated activity from one device or network,
- abnormal timing,
- whether an ad was actually viewable,
- whether a request came through an authorized advertising supply path,
- signs that a browser or device is automated.
IAB Tech Lab’s current Security & Fraud work includes standards such as ads.txt, sellers.json, SupplyChain objects, and ads.cert, all intended to make the programmatic advertising supply chain harder to impersonate or manipulate.
Its Open Measurement SDK also exists partly to verify whether advertising impressions were genuinely viewable and measurable across apps, connected TV, and online video.
Those are legitimate protective functions.
The same observation can support several purposes
Now comes the difficult part.
A signal useful for detecting fraud can also be useful for identifying or profiling a user.
IP addresses can help spot impossible traffic patterns.
Device characteristics can distinguish a real phone from an emulator.
Behavioral timing can help identify automated clicking.
Persistent identifiers can help detect one entity creating thousands of supposedly independent events.
The fact that a signal is useful for security does not mean every later use is automatically justified by the security purpose.
That is the boundary worth inspecting.
The industry is already trying to separate anti-fraud from identity
IAB Tech Lab’s ads.cert roadmap is revealing here. Its work on authenticated devices describes a goal of allowing devices to attest that requests are legitimate in ways that help with invalid-traffic and anti-fraud efforts without enabling user tracking. See IAB Tech Lab’s ads.cert documentation.
Its newer ID-Less Solutions Guidance likewise discusses how advertising systems can perform functions such as measurement, frequency control, and fraud detection in environments where conventional identifiers are unavailable.
That matters because it separates two questions that are often lazily merged:
Do we need signals to detect fraud?
and
Do we need a persistent behavioral identity for that purpose?
The answer is not always the same.
Purpose limitation is the useful audit
A fraud-detection system should be evaluated by what it actually collects and what happens next.
Useful questions include:
- Is the signal retained only as long as needed for fraud analysis?
- Is it reused for ad targeting or audience enrichment?
- Is it shared with unrelated recipients?
- Can the same protective result be achieved with less identifying data?
- Are aggregated or privacy-preserving signals available instead?
The existence of fraud is not a fictional excuse.
It is a real engineering problem.
But a real protective purpose does not turn every observation into a blank check.
The Surveillance Economy becomes harder to analyze when all tracking is treated as morally identical.
The better question is narrower:
What was this signal collected to accomplish, and did the system stop there?