HTTP/2 connection fingerprint (Akamai)
HTTP/2 settings, window updates and header order differ by browser and are readable by the server.
Akamai's fingerprint (Black Hat EU 2017) joins SETTINGS values, WINDOW_UPDATE, PRIORITY frames and pseudo-header order. Chrome sends m,a,s,p, Firefox m,p,a,s and Safari m,s,p,a; Go and Python clients differ visibly.
What it can and can't do
What the user does: Nothing. It happens as your browser connects to the site
It can
- Compare connection settings and header order with known browsers
- Tell Chrome, Firefox and Safari apart from each other
- Spot scripts and bots whose settings look different
It cannot
- Know your name or account from it
- Tell two people on the same browser apart
- Apply to old HTTP/1.1 connections
Browser support
How a site asks for it
| API | HTTP/2 frame capture at the edge |
| How access is granted | none |
| What the user sees | Not listed |
| Needs a secure context (HTTPS) | No |
| Needs a user gesture | No |
| How long it lasts | Not listed |
| How to take it back | Not listed |
| Inside an iframe | Not listed |
Typical flow
- Capture first HTTP/2 frames
- Format SETTINGS|WINDOW_UPDATE|PRIORITY|order
- Compare to known list
Data you get
What comes back: Pipe separated string
Shape: string
| Field | Type | Unit | Range | Precision |
|---|---|---|---|---|
akamai_h2 | string | None | SETTINGS|WINDOW|PRIORITY|ORDER | exact |
Technical limits
- Spoofable by custom HTTP stacks
- Same fingerprint across browser versions sharing an engine
The story
What it is
The way your browser opens a modern web connection leaves a signature the site can read.
For you
You never see it. When you connect, your browser sends settings in a pattern that is almost the same for every copy of Chrome, Safari or Firefox. A site can tell the real thing from a lookalike script.
For makers
Security vendors like Akamai and Cloudflare use it to tell real browsers from bots that fake a browser name. It catches many scrapers and card testing tools. It also adds to a fingerprint that is hard to change.
Where you see it
| Type | Example | Apps |
|---|---|---|
| security | A bot defense compares the connection pattern with the claimed browser and challenges mismatches. | Akamai, Cloudflare |
| shopping | A retailer stops price scrapers that pretend to be Chrome. | None named |
| games | A game launcher site blocks automated tools that buy limited items. | None named |
| ads | An ad platform removes fake traffic from automation tools. | None named |
| banking | A bank adds the connection pattern to its fraud score at login. | None named |
| analytics | A traffic analyst uses it to name the real browser share. | None named |
Risks
- Hard to change
- Can flag privacy tools
Good uses and abuse
Legitimate uses
- Bot detection at CDNs
Known abuse
- Cross-session client identification
Scores
| Score | Value | Meaning |
|---|---|---|
| Sensitivity | 2 of 5 | How revealing the data is |
| Utility | 4 of 5 | How useful it is to honest apps |
| Friction | 0 of 5 | How much the user has to do |
| Fingerprint bits | 3 | Rough identifying entropy, when known |
Related capabilities
Related
- HTTP header order and casing. Can tell which browser or tool you use from the order of request headers
- JA4+ network fingerprint family. Can recognize your browser or app from several quiet network clues, even when shuffled
Sources
- Shuster et al., Passive Fingerprinting of HTTP/2 Clients (Black Hat EU 2017)
- lwthiker/curl-impersonate
How well checked: Based on official documentation.