Documentation
Stepwise Secret & PII Scanner for Confluence: user guide
Everything the app does, in the order you will meet it: the rules, what is checked, old versions, the history scan and the details admins ask about.
On this page
Getting started
Stepwise Secret & PII Scanner for Confluence is a Confluence Cloud app. A Confluence admin installs it from the Atlassian Marketplace and approves its permissions (they are explained under Permissions). It is a separate app from Stepwise Secret & PII Scanner for Jira.
Open the scanner
- Open Confluence settings, then Apps, then Stepwise Scanner. You need to be a Confluence admin.
- The Overview shows open findings by kind, what has been removed or dismissed, and Insights.
- Press Scan existing content to check everything that is already in Confluence. See Scan your history.
From then on, pages, blog posts and comments are checked when they are created or edited. Space admins see their own space under Space settings, Sensitive data.
The app only reads your content. It never edits a page, a comment or a version.
Rules and kinds of data
The Rules tab lists every detector by category. Each can be switched on or off. The app has 42 built-in detectors: pattern matching plus a check where one exists (a checksum, a length, a context). There is no AI and no outside service.
| Detector | Category | Severity | Default |
|---|---|---|---|
| AWS access key, AWS secret key | Secrets | High | On |
| Google API key, Google OAuth client secret | Secrets | High | On |
| Azure storage key, Azure SAS token, Azure client secret | Secrets | High | On |
| GitHub token, GitLab token, Slack token, Slack webhook | Secrets | High | On |
| Stripe live key | Secrets | High | On |
| Stripe test key | Secrets | Low | Off |
| Atlassian, OpenAI, Anthropic, npm, PyPI, SendGrid, Shopify, DigitalOcean, Hugging Face, Docker, Telegram, Postman, Linear, Notion, Figma and Vault tokens | Secrets | High | On |
| Private key (RSA, EC, OpenSSH, PKCS#8, PGP) | Secrets | High | On |
| JSON Web Token | Secrets | Medium | On |
Password written as password: …, in links, configs and sentences, in 20 languages | Secrets | High | On |
Secret in a config line (api_key = …, client_secret: …) | Secrets | High | On |
Credentials in a link (user:password@host) | Secrets | High | On |
| Authorization header (Bearer, Basic) | Secrets | High | On |
| Card number (Luhn-checked) | Financial data | High | On |
| IBAN (checksum for its country) | Financial data | Medium | On |
| US Social Security number | Personal data | High | On |
| Email address | Personal data | Low | Off |
| Phone number | Personal data | Low | Off |
| UK National Insurance number | Personal data | Medium | Off |
| Spanish DNI / NIE | Personal data | Medium | Off |
A match covers only the secret part. For password=hunter2x it is the value, not the word password. When matches of two detectors overlap, the more specific one is kept.
What is not a finding
- Placeholders:
<password>,{{token}},${SECRET},%PASS%,$VAR,***,xxxx,REDACTED,null,none. - References to a vault or an environment:
vault:,env:,process.env.. - Sentences such as “my password is expired”: a spoken password has to look like a password.
- Well-known test data, while Ignore well-known test data is on (the default): documentation keys such as AWS
…EXAMPLE, test card numbers, example IBANs and SSNs.
Ignore values and exclude spaces
- Ignore values containing: up to 50 values of up to 100 characters, for example
example.comor a test key prefix. Case does not matter. - Excluded spaces: up to 200. Nothing in them is checked.
- Dismiss a finding with a reason: Not sensitive, Test data or Accepted risk. It stays dismissed while the match keeps its length. A different length opens it again.
What is checked, and when
- Pages: the title and the body, including code blocks and the parameters of macros.
- Blog posts, the same way.
- Comments: footer and inline.
The app checks a page, blog post or comment when Confluence reports that it was created or changed. A check is not instant: Confluence delivers the event to the app after a short delay. Some comments come without an event, for example an inline comment added by a script through the REST API, so the app also looks at comments changed in the last minutes every five minutes: such a comment is checked within about five minutes.
Life of a finding
- Open
- A match is in the text now.
- Removed
- The value is gone: someone edited it out, the comment or page was deleted, or an admin marked it.
- Dismissed
- A person decided it is not a problem.
Checking the same text again changes nothing. If a value comes back after it was removed, the finding opens again.
Findings and Preview
The Findings tab lists findings with filters for space, state, category and kind. Each row shows the kind, the page or comment, when it was found and its state.
- Preview reads the text as you, finds the match and shows about 40 characters around it with every found value hidden. For cards and SSNs you see the last four digits; for tokens with a public prefix, the prefix. Nothing is saved.
- Dismiss, Reopen, Mark resolved and Check old versions act on the finding.
There is no masking in Confluence. Edit the page and remove the value yourself; the finding tells you what to do.
Old versions
Confluence keeps every version of a page. Removing a password from the current page leaves it in the earlier versions, and anyone who can view the page can open them in Page history.
- When a finding appears on a page or blog post, or becomes removed, the scanner checks its last 200 versions in the background, at most once every 10 minutes per page, and remembers which versions held which kind of data. It stores version numbers and counts only.
- Press Check old versions on a finding to do it on demand, with your permissions.
- The finding then says, for example: Old versions 3, 4 and 7 still contain an AWS access key. Anyone who can view this page can open them in Page history. Delete these versions, then rotate the key.
What to do
- Rotate the credential first: treat it as exposed.
- Delete the listed versions in the page’s version history (More actions ••• → Version history, then Delete in each version’s ••• menu), if your Confluence admin allows it.
- Press Check old versions again to see that the list is empty.
The app does not delete versions. It has no permission to change content, so the deleting is a step you take in Confluence.
Scan your history
On the Overview, Scan existing content checks every space (except the ones you excluded): pages and blog posts with their text, then comments.
- It runs in the background in small chunks, so a large site is fine. Progress shows spaces, items and findings so far.
- You can pause, resume or stop it. If it stalls, it restarts by itself within minutes.
- It waits when Confluence asks it to slow down.
- Findings from the scan are listed like any other.
Insights and the space page
The Overview shows findings by space (top 10), by kind (top 10) and by week (last 8 weeks). Space admins open Sensitive data in Space settings for the findings and Insights of their space. The rules there are read-only.
Journal and CSV
The Journal tab records what happened, where and when. It never holds the value.
- Found, reopened, removed and dismissed.
- Scan started, finished or stopped, and settings changed.
- Who: only app or user, never a name or an account ID.
- Entries are kept for 13 months. Removed findings are kept for 180 days.
Permissions
Who can do what
| Action | Who |
|---|---|
| Open the scanner, change rules, run the history scan | Confluence admins |
| See findings and Insights of a space | Space admins, in Space settings |
| Preview, Dismiss, Check old versions | Admins; Preview shows only what that person may see in Confluence |
What the app is allowed to do
When the app is installed, Confluence asks the admin to approve six permissions. Five are read-only and one is storage. The app has no permission to write to Confluence.
| Permission | Why the app needs it |
|---|---|
read:confluence-content.summary | To be told when a page, blog post or comment is created or edited |
read:page:confluence | To read pages and blog posts with their text and old versions |
read:comment:confluence | To read footer and inline comments |
read:space:confluence | To list spaces and to check whether the current person is a space admin |
read:confluence-user | To check whether the current person is a Confluence admin |
storage:app | To keep settings, findings, the journal and the scan state in Atlassian-hosted app storage |
The app acts as itself, not as a person, where nobody is present to act: reading new text when Confluence reports an event, the history scan and the old-versions check. What you do on screen runs with your own permissions.
Limits
| Item | Limit |
|---|---|
| Detectors | 42 built-in |
| Custom patterns (Premium) | 20, up to 200 characters each |
| Values to ignore | 50, up to 100 characters each |
| Excluded spaces | 200 |
| Text checked per page or comment | 2,000,000 characters (longer text is marked as checked partly) |
| Findings per kind in one place | 100 |
| Old versions checked per page | The last 200 |
| Old-versions check | At most once every 10 minutes per page |
Data and privacy
- Settings, findings, the journal and the scan state are stored in Atlassian-hosted app storage (Forge storage), separately for your site.
- A finding holds: the space, the page or comment, the kind, severity, state, how long the match is, its line and column, the page version, when it was found and last seen, and who changed its state (app or user). For old versions the app keeps version numbers and counts per kind.
- Not stored: the found value, a hash of it, any text around it, page titles, names, emails and account IDs.
- Nothing is written to pages, comments or properties. The app has no write permission.
- Nothing is sent to any other server. There are no external services, no analytics and no AI. Data follows your Confluence data residency.
More detail: Security and data. Legal wording: Privacy policy.
FAQ
- What does Stepwise Secret & PII Scanner for Confluence find?
- AWS, Google Cloud and Azure keys; GitHub, GitLab, Slack, Stripe, Atlassian and OpenAI tokens; private keys; JWTs; passwords written in text and configs; credentials in links; card numbers (Luhn-checked); IBANs; and US Social Security numbers. Email addresses, phone numbers, UK NINO and Spanish DNI can be switched on. 42 built-in detectors in all.
- What does it check?
- Pages, page titles, blog posts, and footer and inline comments, as they are created or edited, plus a one-off scan of everything that already exists. Code blocks and macro parameters are included.
- I removed the password from the page. Is it gone?
- Not from Page history. The scanner lists the old versions that still contain it, so you can delete them and rotate the credential. Deleting versions is a step you take in Confluence.
- Does it change or delete my pages?
- No. The app only reads. It has no permission to edit pages, comments or versions.
- Does it store the secrets it finds?
- No. It stores where a finding is (page or comment, line and column), its kind and its state. It does not store the value, a hash of it or any text around it.
- How do I cut false alarms?
- Switch off kinds you do not need, keep “Ignore well-known test data” on, add values to ignore, exclude whole spaces, and dismiss a finding with a reason.
- Who can see the findings?
- Confluence admins see the whole site. Space admins see their space on the Sensitive data page.
- Does it use AI?
- No. Detectors are patterns with checks such as a checksum, and everything runs inside Atlassian.
- Does it work with the Jira version?
- They are separate apps. Each is installed on its own. See Stepwise Secret & PII Scanner for Jira.
- Does it support dark theme?
- Yes, it follows each person’s Confluence theme automatically.
Troubleshooting
- I do not see a finding I expected
- Check that the kind is switched on, that the space is not excluded, that the value is not on the ignore list, and that it is not a placeholder or well-known test data. Then press Scan existing content to cover older text.
- Too many findings from test data
- Keep “Ignore well-known test data” on, add the prefix of your test keys to the ignore list, or exclude the test space.
- The old versions list is empty but I expected versions
- The check looks at the last 200 versions and runs at most every 10 minutes per page. Press Check old versions to run it again.
- I removed the value and the finding is still there
- A finding closes when the next check no longer finds the value. Edit the page, wait a moment and open the finding again, or press Mark resolved.
Not available yet
These are not in the app today. We list them so that you know what not to look for yet. There are no dates attached.
- Deleting old versions from the app (you do it in Page history)
- Masking of comments
- Email alerts
- Checking attachments
- Scheduled history scans
Missing something else? Tell us.