Skip to Content
Platform.Ontology

APFC / DNA device fingerprint (abuse prevention features)

Collects about 48 device and browser traits, encrypts them, and attaches them to later requests.

Used to track youGraphics and fingerprintingComposite fingerprintWeb browsersWhen the page loadsWhile you use itWhen you are signed in

In one line

Can collect about 48 device traits and send them in secret with later requests

Collects about 48 device and browser traits, encrypts them, and attaches them to later requests.

Builds a device fingerprint from hardware, display, locale, rendering and storage traits. Because it is encrypted with a LinkedIn key and sent as a header, people and researchers cannot easily see what leaves the browser.

What it can and can't do

What the user does: Nothing. It happens when a LinkedIn page loads

It can

  • Collect about 48 traits about your hardware, screen, language, drawing and storage
  • Lock the result with a LinkedIn key
  • Attach it to later requests as a header, so it is hard to see
  • Link your visits to one device profile

It cannot

  • Ask permission first
  • Be seen or checked by you in plain form
  • See your files or private messages
  • Be cleared by deleting cookies alone

Browser support

BrowserStatusFirst versionNote
ChromeSupportedNot availableNone
EdgeSupportedNot availableNone
FirefoxSupportedNot availableNone
SafariSupportedNot availableNone
Safari on iPhoneSupportedNot availableNone
Chrome on AndroidSupportedNot availableNone
Samsung InternetSupportedNot availableNone

How a site asks for it

APIFingerprint2.js derived collector; RSA encrypt (key apfcDfPK); compressToBase64; globalThis.apfcDf; header on API calls; /platform-telemetry/li/apfcDf, /apfc/collect
How access is grantednone
What the user seesNone (no browser prompt)
Needs a secure context (HTTPS)No
Needs a user gestureNo
How long it lastsCookies li_bapfcc and li_odapfcc 2 years (control how many features are collected); li_apfcdc 10 hours (triggering)
How to take it backNot listed
Inside an iframeNot listed

Typical flow

  1. Collector gathers about 47 keyed features plus the getFeatures meta item (48 in BrowserGate's count)
  2. Each feature has a dataLimit (e.g. webgl 6, fonts 1500, userAgent 1500)
  3. Result is RSA encrypted with a key delivered from apfcDfPK, compressed and stored in globalThis.apfcDf
  4. Attached as an HTTP header to later API requests (LIX flag sync.apfc.headers) and posted to telemetry endpoints
  5. Server side flags (pemberly.tracking.fireApfcEvent, pemberly.tracking.apfc.network.interceptor) switch it on

Data you get

What comes back: Encrypted blob of about 48 features
Shape: object
How often: per page or session

FieldTypeUnitRangePrecision
featuresobjectfeatures47 keys plus getFeaturesnames verified in public script snapshot
apfcDfbinarybytesunknown (encrypted, compressed)opaque to the user

Technical limits

  • Cannot be decrypted by the page or user (private key is server side)
  • Feature count and set may change between releases
  • Public snapshot may differ from production

The story

What it is

LinkedIn's abuse prevention code collects around 48 device and browser traits, encrypts them and sends them with later requests.

For you

You never see it. When you use a site that loads this code, your browser quietly collects about 48 traits and sends them in an encrypted bundle. It is meant to catch fake accounts and scrapers.

For makers

LinkedIn says this is for abuse prevention. Fraud teams use it to spot fake accounts and scraping at scale. Privacy researchers point out that the same data can also work as a durable device ID.

Where you see it

TypeExampleApps
socialA professional network blocks a scraper that creates thousands of fake profiles to collect names and job titles.LinkedIn
securityA site compares your device bundle with past logins and asks for a code if it looks new.None named
adsAn ad system could use a similar bundle to recognize a device, which is why privacy groups watch it.None named
datingA dating app uses a device bundle to ban scammers who keep making new profiles.None named
shoppingA marketplace spots one person running hundreds of seller accounts from the same laptop.None named
bankingA bank adds a device trait bundle to its risk score at login.None named
analyticsA research team compares the traits collected and shows which are most unique.None named

Risks

  • Hard to opt out
  • Encrypted so hard to inspect
  • Doubles as a device ID

Good uses and abuse

Legitimate uses

  • Bot and fake account detection
  • Account takeover defense

Known abuse

  • Cross-session device identification
  • Profiling beyond security needs

Scores

ScoreValueMeaning
Sensitivity4 of 5How revealing the data is
Utility3 of 5How useful it is to honest apps
Friction0 of 5How much the user has to do
Fingerprint bits30Rough identifying entropy, when known

A real example: LinkedIn

Core fingerprint stack. The 47 collector keys plus getFeatures give the 48 named by BrowserGate. Names verified in the public mdp snapshot; the count and that production matches today are from BrowserGate and not independently confirmed.

Sources

How well checked: Checked against real behavior.

On this page