Ban the device, not the email.

Banned users come back with a new address and a fresh browser. A device ban follows them across browsers, private windows and cleared cookies.

SignalscrossBrowser
scenario
Step 4 of 4Different deviceAllow
Figure 1After banning @spammer. Illustrative data. Hover or focus to pause; pick a step to scrub.

How it plays out

The attack, what the verified event shows, and what your code does about it.

01 · The attack

How it usually goes

  1. New emailThe first thing a banned user tries.
  2. Cleared cookiesWipes cookie-based bans.
  3. Another browserGets a new visitorId, but the same deviceId.
02 · What Trace sees

Fields on the verified event

  • deviceIdStable across every browser on the machine.
    "deviceId": "Xk3pQ9vLmT2aR7sBn4Wc"
  • crossBrowserOther browsers that look like this device.
    "linkedVisitors": [  { "browser": "firefox", "confidence": 0.91, "sameNetwork": true }]
03 · What you do

Decide, with your own thresholds

  • AllowDifferent device
  • BlockBanned, device stored
Read the guide

Decision lab

Switch signals on and off. The score is a plain sum of weights.

decision lab
Signals

Weights are the defaults in scoring.default.json. The cut-offs at 25 and 60 are examples; your app sets its own.

15suspectScore / 100
AllowLet the action through.
15/ 100
reasons
  • crossBrowser+15

Implementation

One browser call, one server check.

01IdentifyCall the SDK on the action and send the eventId with it.
02VerifyYour server reads the event with a secret key.
03DecideAllow, challenge or block, with your own thresholds.
identify.ts
const trace = createClient({ apiKey: 'pk_...', endpoint: 'https://id.example.com' });

// On the action (signup, login, checkout), then send eventId with the form.
const { eventId } = await trace.get();
Example response, illustrative
{  "eventId": "evt_example"}

Start in shadow mode. Block later.

Log the decision for a week, check who it would have stopped, then turn it on.