Skip to content

Security

What we do with your code, and what we will not claim about it

A security product that is vague about its own security is not worth trusting with a repository. Every control below says whether it exists in the code today or is still planned, because a page that lists both without distinguishing them is a page claiming security it has not built.

Reporting a vulnerability

Write to security@audixa.com. We will not take legal action over good-faith research, and we will tell you what we did about it.

There is no bounty programme yet. Saying so is better than implying one and disappointing somebody who spent an afternoon on us.

Built today

10 controls that exist in the code now

These are properties of how the application is written, not policies we intend to follow. Each one is enforced somewhere a developer cannot forget it.
Verifiable in the current release
  • Your data is separated by organization, and that separation is tested

    Every query against business data carries the organization it belongs to, enforced by the database layer rather than by each developer remembering. A generated test drives every API route with another organization's identifiers and requires a 'not found' answer.

  • Permissions are checked on the server, on every request

    Roles are evaluated from one shared capability matrix. Hiding a button is not a control, so every page and endpoint re-checks rather than trusting what the browser was shown.

  • Changes to access take effect immediately

    Roles and account status are read from the database on each request instead of being cached in your session, so removing somebody's access removes it now, not when their session expires.

  • Sensitive changes are recorded in an append-only audit log

    Who invited whom, who changed a role, who altered settings. The application has no permission to update or delete those rows; the database rejects the attempt.

  • Secrets are kept out of logs

    Credential-shaped values are redacted before anything reaches a log sink, and the same scrubbing runs independently over audit metadata.

  • We do not run your code

    Dependency analysis reads your lockfile; it never installs or resolves anything. There is no build step, no test run and no package script executed anywhere in a scan.

  • Repository analysis happens in a throwaway sandbox with no network

    Each scan gets its own container with networking disabled, a read-only image, memory and CPU limits and a wall-clock kill. The workspace is destroyed when the job ends.

  • Active checks only run against a domain you have proved you control

    Verification is a DNS TXT record or a file at /.well-known. Until one of those succeeds, we do not send a request to your domain.

  • Our scanner cannot be pointed at a private network

    Every outbound request resolves the hostname first, refuses any address that is not public, and connects to the address it validated — so a DNS answer cannot change between the check and the connection. Redirects are re-validated at every hop.

  • Repository access is read-only and short-lived

    We ask for permission to read contents and metadata, nothing more, and each scan uses a token that expires. We never write to your repository and never open a pull request.

Not built yet

0 controls the scanning milestones will add

These describe how scanning will work. The scan engine is not in the current release, so nothing here is a claim about today — it is what we have designed and what we will have to prove.
Planned, with the milestone named

    Limits

    What a scan is not

    The most useful section on this page for deciding whether Audixa fits what you need.
    • A clean scan means the checks we ran found nothing — not that your application is secure. We tell you which checks ran and which did not.
    • We do not attempt to exploit anything. No brute forcing, no port scanning, no destructive payloads.
    • We do not review business logic that only a human who knows your product could evaluate.
    • We are not a compliance certification. We do not issue ISO, SOC 2 or PCI attestations, and a scan is not an audit.
    • Automated analysis finds patterns. It misses things, and it sometimes flags something that turns out to be fine — which is why every finding carries a confidence level and its evidence.

    We will never tell you your application is secure, and we will not write a letter saying it. The best an automated review can honestly say is which checks ran and what they found — which is genuinely useful, and is what every scan report says.

    Data handling

    What we store, and for how long

    Retention, the providers that can see anything, and how to have your records deleted are all on the privacy page, written as tables rather than as paragraphs of reassurance.
    Read the privacy notice

    The short version

    Your source code is not retained. A scan works in a container that is destroyed when the job ends, and what survives is the findings plus up to three lines of context around each match, with anything that looks like a secret redacted before it is stored.

    Enquiries sent through this site are kept for 24 months from the last contact, and we will delete one sooner if you ask.

    Read the whole thing before connecting anything

    If a control you need is in the planned list rather than the built one, that is worth knowing now. Tell us which one and we will say honestly when it lands.