An embedded form or survey builder still ships skip logic and scoring in the browser
Form and survey builders often embed a runtime in the customer's page: a script tag, a webpack chunk, or a Web Component. The pitch is instant skip logic, live validation, calculated scores, and no round-trip between questions.
That runtime is JavaScript, and it usually includes the definition of the form itself. If the value of your product is the engine - branching, scoring, price quotes, qualification - then that engine is in the browser.
This article is about what an embedded form runtime actually exposes, what must be re-checked on the server, and how to protect the engine you still have to ship.
The definition and the engine travel together
A typical embed looks like this:
export function mountSurvey(el, definition) {
let answers = {};
function next() {
const q = pickNext(definition, answers); // your interesting part
if (!q) return finish(score(definition, answers));
renderQuestion(el, q);
}
el.addEventListener("submit", (e) => {
answers[e.qId] = e.value;
next();
});
next();
}
function pickNext(definition, answers) {
// skip logic, AB branches, "disqualify if β¦", paid-tier questions
return definition.rules.find((rule) => match(rule, answers));
}
function score(definition, answers) {
return definition.weights.reduce((n, w) => n + w.points * !!answers[w.id], 0);
}
A visitor who loads the embed can:
- Read the runtime chunk and the form definition from the Network panel.
- Recover skip rules, hidden questions, disqualification thresholds, and scoring weights.
- Call
scoreorpickNextfrom the console with crafted answers. - Lift the engine into another product, or fill the official form with a script that already knows the winning path.
Minifying the runtime will rename pickNext. It will not hide "disqualify", a distinctive weight, or a question you thought was secret because the UI had not reached it yet.
grep -R -nE "disqualify|weights|pickNext" dist/embed | head
Hidden questions are not confidential
This is the part teams regret after a launch. A hiring screen, a pricing quiz, a medical-style symptom tree, or a lead-qualification form sometimes puts the "real" criteria in client-side skip logic: reject if salary is below a number, skip the expensive plan if the company size is small, show a deluxe upsell only when the score is high.
Assume anyone can read those criteria. If they must stay unpublished - interview answer keys, fraud signals, medical cutoffs, unpublished prices - do not put them in the embed. Fetch the next question from the server, or at least re-evaluate the result on the server before it means anything.
Submitted answers should also go to your server in raw form. A client-computed score: 97 is a suggestion.
What the embed should still do
A snappy runtime is still the right UX for most marketing forms. Validation, animations, save-and-resume, and non-sensitive branching can stay in the browser.
For the proprietary engine:
- Re-check qualification, prices, and pass/fail on the server before you create a lead, a charge, or a "you passed" state.
- Minify the embed as usual.
- Extract
pickNext/score(or the compiler that turns a form JSON into those functions) and apply a real protection pass - identifier renaming and string tables as a floor, VM-style protection if wholesale reuse of the engine is the risk you care about. - Keep unpublished answer keys and fraud signals off the client.
- Keep production source maps off the public CDN.
A JS VM obfuscator compiles that function to a custom bytecode and ships a small interpreter. Pretty-printing the chunk no longer reconstructs the original scoring routine; you get bytecode and a dispatcher. That is slower than a thin JSON walker, which is why you protect the engine, not the entire form UI.
VEILVM is a browser-based JSVMP workspace you can point at one extracted function before you wire protection into CI. Protection raises the cost of casual inspection and wholesale reuse. It does not make a public form definition a secret, and it does not replace a server-side score.
Checklist
- [ ] Qualification, prices, and pass/fail are re-evaluated on the server. A client-computed score cannot grant the result by itself.
- [ ] Unpublished keys, fraud signals, and confidential cutoffs are not in the embed.
- [ ] The open UI kit is left alone; only your skip or scoring engine is a candidate for protection.
- [ ] Distinctive rule strings and weights in the public chunk are ones you accept being public, or they live behind a stronger pass.
- [ ] Production source maps are not downloadable from the CDN.
- [ ] End-to-end tests submit a fixture form against the minified (and, if applicable, protected) production embed, including a path the UI would not show.
If you want to try VM protection on a single extracted function, VEILVM runs in the browser with no sign-up: 3 free runs a day, up to 5 KB per file (resets at 00:00 UTC). A free account raises the input limit to 30 KB. Pro is $9/month or $79/year and adds an API so the protection step can run in CI. Standard requests are processed in memory and are not written to logs. Start at https://vm.ldvmp.com.
Comments
No comments yet. Start the discussion.