Essay

Why fraud models should only propose. By Amir Shachar.

Read the essay
Skip to content
Riskernel

Free tool · Runs in your browser

Launch attack-tree generator

Choose what you are launching and tick the controls you already run. You get a tree of the ways attackers go after it, each branch ending in concrete steps, and every step is marked covered or open. Export it to share with your team.

Everything is worked out in your browser. Nothing you tick is sent or stored. The link button only builds an address that carries your choices, and the download buttons make files on your own device.

How this is calculated

The tree comes from a fixed library that we wrote and reviewed by hand. Each product type has one goal for the attacker at the root, several attacker paths below it, and concrete steps at the ends, which we call leaves. Each leaf is a well-known fraud pattern that pays a real attacker, described at the level of a conference talk: what the attacker does and why it pays, never how to do it.

Each leaf also carries a short list of the controls that can address it. Ticking a control only changes how the leaves are marked. The rule is the same for every leaf:

covered   at least one ticked control is on the leaf’s list
open      no ticked control is on the leaf’s list

summary   open leaves of all leaves

The summary line counts the open leaves, for example 9 of 14 leaves have no listed control. The checklist only shows controls that matter for the product you picked.

Why open and covered are a starting point

A control on a leaf’s list means it can address that pattern. It does not tell you whether your version is strong enough, sits at the right point in the flow, or is tuned for your traffic. A covered leaf is a claim to test. An open leaf is a question to ask. The library also cannot see paths that are specific to your product.

Frequently asked questions

Does the generator send my choices anywhere?

No. The tree is built in your browser from a fixed library on this page. The download buttons make files on your own device, and the “Copy link” button puts your choices in the page address so you can share them, and nothing else.

Is this a full threat model?

No. It is a starting map for a launch conversation. It covers common attacker paths for six product types and does not see your actual flow, limits, vendors or data. The Launch Fraud Review goes through your real flow and writes down what we find.

How was the library built?

We wrote a fixed library of well-known fraud patterns for each product type, each one a pattern that pays a real attacker, and mapped each to the controls that can address it. A person reviewed every entry. Nothing is generated when you open the page, so the same choices always give the same tree.

What does covered mean?

A leaf is covered when at least one control you ticked is on that leaf’s list. That says the control can address the pattern. It does not say your version of the control is strong enough, placed at the right step, or tuned for your traffic. Treat open leaves as questions to ask and covered leaves as claims to test.

Can I put the generator on my own site?

Yes. Copy the embed code on this page. The embedded version keeps one credit link back to this page.

Use this generator on your own page

Paste this snippet into a page you control. The embedded version has no site menu and one credit link back here.

Go deeper

This tree is a map of common paths. The Launch Fraud Review takes your actual flow, your limits and your vendors, and writes down which paths are open, how likely each is to pay an attacker, and what to change before launch. Get a Launch Fraud Review