Security and vulnerability disclosure
How to report a vulnerability and what to expect from us.
Last updated
On this page
Version 1.0, effective 7 October 2026.
Reporting a vulnerability
If you believe you have found a security weakness in Cleared-pay, email legal@cleared-pay.com. It is the only published address for security reports today; a dedicated mailbox will be announced here. We do not yet publish a PGP key. If your report contains sensitive detail, send a first message without it and ask for a secure channel.
A report we can act on includes the URL, endpoint or component affected; the steps to reproduce, with requests and responses where relevant; the impact; the value of the x-request-id response header if you have one; and the date and time of your tests. A proof of concept helps. Tell us how you want to be credited, if at all.
Scope
- The website at cleared-pay.com, including the documentation and blog.
- The merchant portal under /portal and the backoffice under /admin.
- The public API under /api and the webhook delivery system.
- The hosted checkout page under /pay and the sandbox authentication page under /3ds-test.
- Email we send from cleared-pay.com.
Use test mode and your own test merchant account; test card numbers are in the documentation.
Out of scope
We will close the following without action, and they do not qualify for credit.
- Reports from an automated scanner with no demonstrated impact, version banners, generic missing-header findings, and best-practice suggestions without a security consequence.
- Clickjacking on pages with no state-changing action, and findings that need a browser with its security features disabled, a compromised device, a malicious extension or an already stolen credential.
- Social engineering of our staff, merchants or vendors, and physical attacks.
- Denial of service, including volumetric testing against the rate limits.
- Findings in a third-party service we use that are not specific to our configuration; report those to the vendor.
What we ask
- Test only against accounts and data that are yours. If you encounter someone else's data, stop, record only what shows the issue, report it, and delete what you hold.
- Do not run denial-of-service tests or spam forms; keep automated traffic within the published rate limits.
- Do not use social engineering, phishing or physical access against anyone, do not modify or delete data that is not yours, and do not pivot from a finding into further systems.
- Give us 90 days from your report before you publish, or less by agreement once a fix is deployed.
What you can expect
- An acknowledgement within three business days.
- A triage decision with a severity within ten business days, or a request for more detail if we cannot reproduce the issue.
- Updates at least every two weeks while the fix is in progress, and a message when it is deployed.
- Credit by name or handle on this page if you want it, once the fix is live, and a direct answer with the reason if we decide not to fix something. We do not run a paid bounty programme.
Safe harbour
If you follow this policy in good faith, we will not pursue legal action against you for your research or refer you to law enforcement for it, and we will work with you to resolve the issue. We consider research conducted in line with this policy to be authorised access. The commitment does not extend to actions outside the policy and cannot bind third parties: on a vendor's system, that vendor's rules apply.
Security practices
The current summary, including encryption, access control, logging and testing cadence, is on the Trust Center. What follows is drawn from the running code: what is in place today, and what is not.
Transport and browser protection
Every response carries a Content-Security-Policy generated per request: scripts run only with a server-issued nonce and strict-dynamic, object-src is none, base-uri and form-action are limited to our origin, and pages cannot be framed. On the hosted checkout page, styles are nonce-only as well. HTTP Strict Transport Security is set for two years with includeSubDomains and preload. Responses also carry X-Content-Type-Options: nosniff, a strict-origin-when-cross-origin referrer policy, and a Permissions-Policy that disables the camera, microphone, geolocation and similar features.
Accounts and sessions
Passwords are at least 12 characters and are hashed with Argon2id (64 MiB of memory, 3 iterations). Two-factor authentication is required: a time-based one-time code (RFC 6238, 30-second steps, 6 digits) plus ten single-use backup codes; the secret is stored AES-256-GCM encrypted, the backup codes as Argon2id hashes, and a code is accepted once per time step. Creating or revealing an API key, changing a payout bank account, a refund above the merchant's threshold and every backoffice write require a fresh code within the last five minutes. Portal sessions end after 30 minutes without activity or 12 hours in total; backoffice sessions after 15 minutes or 8 hours. A password or role change revokes every other session. Sign-in attempts are limited to 10 per 15 minutes.
Data isolation and encryption
Each merchant's rows are isolated by PostgreSQL row-level security: the application connects as a role that sees only the tenant set for the current transaction, and the database enforces the policy, not application code. Bank account numbers, two-factor secrets and webhook signing secrets are encrypted at the column level with AES-256-GCM, bound to their row so a ciphertext cannot be moved between records, in a versioned format so the key can be rotated. Onboarding documents and dispute evidence are private objects, encrypted at rest, reached only through signed URLs valid for 15 minutes.
API keys, webhooks and rate limits
An API key is shown once; we store a lookup prefix, the last four characters and an HMAC-SHA256 of the key under a secret pepper, with different peppers for test and live so a leak of one cannot expose the other. A rolled key keeps working for a grace window of 24 hours by default and 7 days at most. Webhooks are signed with HMAC-SHA256 over the timestamp and the raw body in the Cleared-Signature header; a signature older than 300 seconds is rejected, and a secret can be rotated with both old and new valid during the window. Rate limits per merchant are 100 payment creations, 300 writes and 1,000 reads per minute and 10 API key creations per hour, enforced atomically in the database.
Card data
Card numbers never reach a merchant's systems and are never stored by us. On the hosted page the card fields post to our own origin, are tokenised within that request and discarded; the API rejects a raw card number in a request body. What remains is the brand, last four digits, issuing range, expiry, issuer country and a keyed fingerprint. A redaction layer removes keys, secrets, authorisation headers and any card-number-shaped value before a log line is written. The checkout page carries a documented script inventory for PCI DSS requirement 6.4.3; no third-party script runs there, and the end-to-end suite fails if one appears.
Audit and change control
Every write to a merchant-owned record produces an audit entry in the same database transaction, with the actor, request identifier, IP address and user agent. The audit log, payment event log and API request log are append-only: the application's database role cannot update or delete them. Changes reach production only from version control, after type checks, lint, tests, a contrast check, a content check and a production build; migrations apply only after that pipeline passes.
Hosting
The application, API and hosted checkout run on Vercel in Frankfurt; the database is PostgreSQL on Neon in Frankfurt with seven days of point-in-time recovery; files are in Cloudflare R2 under the European Union jurisdiction setting; email goes through Resend; the data region of that service is confirmed here once its agreement is countersigned.
Not yet in place
- A third-party penetration test of the API, hosted checkout, portal and backoffice is scheduled and not yet performed; the date and scope will be stated here once it has.
- Our own PCI DSS attestation is pending and is published here when issued. Live card processing runs through acquiring partners that hold their own PCI DSS validation; the hosted page is built so that card data is tokenised in the request that receives it and is never stored.
- Tamper detection on the checkout page under PCI DSS requirement 11.6.1 is not implemented; the content security policy blocks unauthorised scripts, and a violation-reporting endpoint is planned.
- An ISO/IEC 27001 certification has not been sought.
- A dedicated security mailbox and a published PGP key are being set up.