Security
Security at Signpost
Last updated October 5, 2026
Video stays on your device
Hand tracking, face tracking, and sign recognition all run in your browser using MediaPipe and TensorFlow.js. Video frames are processed in memory and then discarded. They’re never recorded, uploaded, or stored, and neither are the hand and face landmarks the models produce. Our servers only receive lesson results: which signs you practiced and how accurate they were.
Infrastructure
- Signpost’s web servers are run by our team and sit behind Cloudflare, which handles TLS, filters malicious traffic, and absorbs denial-of-service attacks.
- Account and learning data is stored in a managed Postgres database (Neon) in the United States. It’s encrypted at rest, and connections to it are encrypted too.
- Sign-in is handled by Clerk, so we never see or store your password. Card payments go through Stripe Checkout, so card numbers never touch our servers.
- Our LMS integration runs as a separate service with its own signing keys.
The full list of providers that handle personal data for us is on our Subprocessors page.
Protections in the app
HTTPS everywhere
Every connection uses TLS. HSTS, including subdomains, tells browsers never to connect to Signpost over plain HTTP.
Security headers
Both sites block framing by other sites, prevent MIME sniffing, set a strict referrer policy, and switch off browser features they don't use. The website also sends a strict Content Security Policy and cross-origin isolation headers.
Server-side scoring
XP and progress are calculated on our servers, not taken from the browser. Each lesson starts with a single-use token, so a forged request can't award credit.
Permission checks
Classes, assignments, and admin tools check who you are and what you're allowed to see on the server, on every request.
Rate limits and bot checks
Sign-ups, bug reports, settings changes, and model downloads are rate-limited. The waitlist form also has a hidden bot trap and rejects stale or replayed submissions.
Encrypted model delivery
Recognition models are sent to your browser encrypted with a fresh key for each request (ECDH with AES-256-GCM), and they're never cached.
Signed LMS launches
LMS launches use LTI 1.3, so each one is a signed token we verify against the LMS's published keys. Our LMS service and the app sign every request they send each other.
Safe roster imports
When a teacher imports a Canvas roster, we refuse private and internal network addresses and pin the connection to the address we checked. This blocks server-side request forgery.
Access to data
Only a small number of people on our team can access production data, and only when they need it to run Signpost. Admin tools in the app only work for a short allowlist of verified accounts, and we don’t hand out broad administrative access.
Software updates
We watch for security advisories in the open-source packages Signpost depends on and update when a fix affects us. The code for this website is public on GitHub.
Student data
Many of our users are students, some under 18 and some using Signpost through school. For them:
- There are no ads, no data sales, and no advertising profiles.
- We collect what Signpost needs to work: account details, learning progress, and class information.
- Children under 13 need a parent’s or school’s consent, as COPPA requires.
- When a school uses Signpost, we handle student records on the school’s behalf and only for its purposes, in line with FERPA.
- Students, parents, and schools can ask us to delete student data at any time, and we do it within 30 days.
Schools can find more in our Trust Center.
Reporting a vulnerability
If you think you’ve found a security problem in Signpost, email [email protected] with a description, steps to reproduce it, and the impact you think it has.
While you’re looking into it, please:
- Only test with accounts you own or have permission to use. Don’t access, change, or delete other people’s data, and if you come across any, stop and tell us.
- Don’t run denial-of-service tests, send spam, or try social engineering on our team or users.
- Give us a reasonable amount of time to fix the problem before you share it publicly. We ask for 90 days.
We’ll confirm we got your report within 48 hours and update you at least every 5 business days until it’s resolved. If you act in good faith and follow these guidelines, we won’t take legal action against you over your research. We don’t run a paid bug bounty, but we’re glad to credit you publicly if you’d like. Our security.txt has the same contact details.
If something goes wrong
If we learn of a breach that affects your data, we’ll:
- Contain it and stop any further exposure.
- Work out what happened and whose data was involved.
- Tell affected users and schools without undue delay, and notify regulators where the law requires, including EU and UK authorities within 72 hours.
- Fix the cause, and write up what happened and what we changed.
What we don’t do
- We don’t record, upload, or store camera video or images.
- We don’t use face tracking to identify anyone, and we don’t create face templates.
- We don’t sell personal data or share it with advertisers or data brokers.
- We don’t show ads or build advertising profiles.
Contact
Matrix Studios Software
1968 S. Coast Hwy #3479, Laguna Beach, CA 92651
Email: [email protected]
For security reports, please put “Security” in the subject line.


