Read a sample: 1. Evidence Has a Job
A tester can spend an hour working carefully and still produce a finding that no one can use. The technical action may have been valid, the result may have been meaningful, and the tester may remember exactly what happened. But if the record says only “access control broken,” the reviewer must supply the missing facts: which account was used, what was requested, what came back, and why the result mattered.
Evidence exists to answer those questions. In an authorized security test, evidence is the recorded information that helps someone understand and evaluate what happened during the assessment. It may support a finding, qualify one, or show why a suspected issue could not be confirmed. Its job is not to make a report look technical. Its job is to connect a claim to an observable basis.
Useful evidence is therefore more than a screenshot or a tool log. A screenshot can show a page without showing which account was signed in. A log can contain thousands of lines without identifying the one that matters. A note can accurately describe a response but leave out the request that produced it. Each artifact may be genuine, yet still be insufficient for the decision a reviewer needs to make.
This chapter introduces a simple way to think about that problem: distinguish what you observed from what you think it means, and distinguish both from the conclusion you are prepared to report. We will follow one fictional test from a rough note to a finding another person can understand.
Imagine that a reviewer receives a report several days after testing. The reviewer did not watch the tester work and cannot ask the application what it looked like at the time. The system may have changed, a test account may have been disabled, or a temporary condition may no longer exist.
The evidence should give that reviewer a fair chance to answer questions such as:
- What was the tester trying to check?
- What system, account, and relevant conditions were involved?
- What did the tester do within the authorized test?
- What was actually observed?
- How does that observation support—or limit—the stated conclusion?
Different audiences need different levels of detail. A technical owner may need enough context to investigate a request and its response. A manager may need a clear explanation of the affected function and its significance without receiving a copy of sensitive data. A second tester may need to understand whether a result was repeatable or depended on a particular account.
The same evidence does not have to appear in full for every audience. Working notes, supporting artifacts, and the final report serve related but different purposes. The working record preserves enough context for the assessment team to understand the test. The report selects and explains what the intended reader needs. Neither purpose justifies collecting everything simply because it might be useful someday.
In this book, evidence is an assessment term, not a promise that an artifact meets a particular legal standard. Some engagements require formal custody procedures; others need a practical, well-maintained working record. The applicable agreement and instructions determine the requirements. Even when formal procedures are not required, clear records help reviewers understand what an artifact does—and does not—show.
Suppose a client has authorized a test of a fictional internal application called Harbor Documents. The application lets employees view documents assigned to their own accounts. The client has supplied two test accounts and two harmless test documents for the assessment. The rules allow the tester to check whether one test account can access the other test account’s document.
The intended behavior is straightforward: a user should be able to view their own assigned document, but not a document assigned only to the other test account. Because the client has provided test accounts and documents for this purpose, the tester does not need to inspect an employee’s real files to investigate the question.
The tester signs in as test user Cedar, opens Cedar’s assigned document, and notes that the application uses a document identifier in the request. The tester then changes that identifier to the one for a document assigned only to test user Maple. The application returns a successful response containing the title and body of Maple’s test document.
Immediately afterward, the tester writes:
> Access control broken. Cedar can read everyone’s documents.
That note captures the tester’s concern, but it is a poor foundation for a finding. The test involved one account requesting one other test document. “Everyone’s documents” expands the observation beyond what was checked. The note also omits the target, the action, the response, the account relationship, and the fact that the documents were supplied for testing.
A better first record might be:
> While signed in as test user Cedar, I requested Maple’s assigned test document by changing the document identifier in the request. Harbor Documents returned a successful response containing the test document’s title and body. The client-provided account and document assignment list identified the document as assigned to Maple, not Cedar. I stopped after confirming the response.
This version is not yet a complete evidence record. It still needs details such as when the test occurred, the relevant application environment, and a reference to the supporting artifact. But it separates the observed result from the broader claim and tells a future reader what happened.
The distinction matters because a tester’s memory naturally fills…