Does your security report claim an exposure that is not there?

Health-app security review: a logged-in capture was read as anonymous and nearly produced a false exposure, retracted before delivery. What to ask a tester.
Case study

Does your security report claim an exposure that is not there?

The capture came from a logged-in session. The tool still read it as anonymous.

For CISOs, CTOs and heads of QA at health-software companies who buy security testing.

If your security tester works from a browser capture, the export can drop the login it came from, and the report can then name an exposure that is not there. A clinical documentation app (a React front end with Firebase sign-in) asked for a security review. It had 15 public source-map files, which anyone could download to see how the front end works, and a token relay with no origin check. We found no critical or high finding, ruled out five attack classes, and retracted one false alarm before it reached the report.

  • Healthcare software
  • Clinical documentation web app
  • Compliance
  • Clients
  • Reputation
A BetterQA engineer in glasses working at a laptop behind two monitors in the Cluj-Napoca office
The numbers
0 criticalfindings after five attack classes were ruled out
15public source-map files that exposed the whole front end
0critical or high findings in the final report
5attack classes tested and ruled out
1false alarm, retracted before it reached the report
At stake

What was at risk

Compliance

A health app's internals open to anyone

15 public source-map files exposed the whole front end of a clinical app. In a regulated health setting, that shortens the work of anyone looking for a way in.

Clients

A token relay that answers any page

A message handler with no origin check replied to any target origin. A sign-in token can be passed to a page that should not have it.

Reputation

A report that cries wolf

The capture came from a fully logged-in session, but the export had stripped the Authorization header. The tool read it as anonymous and nearly reported an exposure that did not exist.

Rules

Standards and rules in play

  • Patient-facing app, attack classes ruled outCross-tenant access, path traversal, JWT bypass, SQL injection and admin escalation were each tested and ruled out.
  • A report that does not assert more than it sawThe parser now warns that headers may have been stripped when any of four signals is present, and keeps its original message only when none is.

Health software holds special-category personal data under GDPR, and web security testing is commonly organised around the OWASP Top 10 categories. This page describes testing, not a certification of the client.

Method

How we tested it

Test level
System testing of a deployed web application, black-box, from a captured logged-in session.
Test types
Security testing (non-functional), with a negative control for each probe.
Test basis
A browser capture of a logged-in session and the running application.
  • TechniqueError guessingAttack-style probes for cross-tenant access, path traversal, JWT bypass, SQL injection and admin escalation. A bogus source-map path returned 404 as the negative control.
  • TechniqueExperience-based checking of the instrumentThe export had lost its Authorization header. Four signals now stop the tool from saying nobody was logged in.

Terms as in the ISTQB CTFL 4.0 glossary.

Findings

What we found, and what it costs

Finding

15 public source-map files exposed the whole front end

What it costs youAnyone can read how the app is built, which routes exist and how sign-in works. Cheap to remove, useful to an attacker.

Finding

A token relay with no origin check

What it costs youIt replies to any target origin, so another page can ask for a sign-in token.

Caught

A false report of an unauthenticated exposure

What it costs youReporting something that is not there costs the client engineering time and trust in the whole report. It was retracted before delivery.

The final report had 0 critical, 0 high, 2 medium, 3 low and 3 informational findings.

Outcome

What changed

The report went out with no critical or high finding. The parser now checks four signals before it says nobody was logged in: a referer on a non-login path, personal-data-shaped keys in a response, a 200 on a protected-looking path, and an authentication-provider host.

What this means for you: If you buy security testing for a health app, ask for the list of what was ruled out and what the tester could not see, not only the findings.
Check

What to check in your own product

  1. Ask how your tester knows whether a capture was made logged in or anonymous.The capture came from a fully logged-in session, but the export had stripped the Authorization header.
  2. Ask whether each attack probe had a negative control.A bogus source-map path returned 404 as the negative control.
  3. Check whether your build ships public source-map files.Anyone can read how the app is built, which routes exist and how sign-in works.

Can your security tester show what they ruled out?

Our reports list what held, what we could not see, and what we retracted. The first month is a trial.

Talk to our team

Automate your security testing

Our AI Security Toolkit runs 20 checks we wrote ourselves alongside 15 open-source scanners to find vulnerabilities other scanners miss. SAST, DAST, SCA, secrets detection - unified in one scan.

Explore AI Security Toolkit
Share the Post: