Security Scan Baselines and Drift Detection: Separating New Findings From Known Risk
security baselinesdrift detectionCI/CDregression testingvulnerability scanningrelease security

Security Scan Baselines and Drift Detection: Separating New Findings From Known Risk

SScan Quest Editorial Team
2026-11-08
7 min read

A practical checklist for creating security scan baselines, detecting drift, managing exceptions, and separating new findings from known risk.

A security scan baseline helps a team distinguish a newly introduced regression from a finding that is already known, accepted, or still being remediated. This guide provides a practical workflow for creating a CI/CD vulnerability baseline, detecting drift, reviewing exceptions, and deciding when a baseline should be updated rather than reset.

Overview

A baseline is a recorded view of security findings at a defined point in time. It can include the finding identifier, affected repository or service, file or endpoint, severity, status, owner, and the scan or release that established the record. Later scans compare their results with that view so that developers can focus on changes instead of receiving the same alert on every pipeline run.

Baselines are useful for SAST, SCA, DAST, API security scanning, container security scanning, and cloud security checks. They are particularly valuable when a team is introducing automated security scanning to an existing application with known technical debt. Without a baseline, a pipeline may report hundreds of existing findings and make it difficult to identify a new issue. With a poorly managed baseline, however, a real regression can be hidden.

The central rule is simple: baseline known risk, not future risk. A finding should enter a baseline only after the team has confirmed what it is, where it exists, who owns it, and why it is not being treated as a release-blocking issue at that moment. A baseline is not a permanent exemption and should not be used to make a failing scan appear healthy.

Keep two concepts separate:

  • Finding state: open, confirmed, fixed, accepted, false positive, or under review.
  • Comparison state: new, previously known, changed, disappeared, or reappeared.

This distinction makes vulnerability scan drift detection more reliable. A previously known high-risk issue may remain open, but it should still be visible in risk reporting. A finding that disappears from one scan and returns in another should not automatically be treated as new or permanently resolved.

Checklist by scenario

Creating an initial baseline for an existing application

  1. Define the scan scope, including repositories, branches, services, environments, APIs, images, and dependency manifests.
  2. Run the scan using the same configuration intended for routine CI/CD security scanning.
  3. Normalize duplicate results where the scanner identifies the same underlying issue in multiple locations.
  4. Validate high and critical findings first, then review lower-severity findings according to business risk.
  5. Assign an owner, target review date, and treatment decision to every finding added to the baseline.
  6. Record the commit, release, image digest, scan configuration, and timestamp used to create the baseline.
  7. Store the baseline where security and engineering teams can review its history and permissions.

Do not baseline a result merely because it is inconvenient to fix. If a finding is a false positive, document the evidence and mark it as such through the scanner's supported workflow. If it is accepted risk, record the business context, compensating controls, expiration date, and approving role.

Reviewing a pull request or feature branch

  1. Compare the branch scan with the correct target branch or release baseline.
  2. Classify findings as new, modified, inherited, resolved, or reintroduced.
  3. Prioritize findings introduced by changed files, dependencies, infrastructure definitions, endpoints, or container layers.
  4. Check whether a new result is actually a moved or duplicated instance of an existing issue.
  5. Apply the repository's gate policy consistently. For example, block on defined severity or exploitability thresholds rather than on every historical finding.
  6. Leave a clear review comment that identifies the finding, expected action, owner, and exception path.

Branch baselines should normally be temporary. They help developers work with the changes in a branch, but the release or protected default branch should remain the authoritative comparison point.

Preparing a release

  1. Run the full intended scan set against the release candidate, not only the files changed in the latest commit.
  2. Compare results with the last approved release and the current risk register.
  3. Investigate all new findings, including those that appear outside the immediate feature scope.
  4. Confirm that accepted findings have not exceeded their review or expiration date.
  5. Check for reintroduced findings that were marked fixed in an earlier release.
  6. Record the release decision, unresolved exceptions, owners, and follow-up dates.

A release baseline should describe what was actually shipped. It should not be replaced simply because a scan produced noisy results. If the scan configuration changed, preserve the old record and create a documented transition to the new one.

After a major tool or workflow change

When scanner rules, authentication, scope, severity mapping, dependency resolution, or build inputs change, expect result counts to change. First run the old and new configurations in parallel where practical. Compare a sample of findings and document whether differences reflect improved coverage, changed classification, duplicate detection, or genuine application drift. Create a new baseline only after the comparison has been reviewed and approved.

What to double-check

Before relying on a security scan baseline, verify the following:

  • Scope: Does the scan cover the same assets each time? A missing repository, API route, image, or authenticated user flow can create a false impression of improvement.
  • Identity: Can the tool recognize a finding when its file, line number, endpoint, package version, or container layer changes? Stable identifiers are preferable to location alone.
  • Configuration: Are rules, exclusions, severity mappings, credentials, and environment variables recorded with the baseline?
  • Ownership: Does each accepted or open finding have an accountable team rather than a generic queue?
  • Expiration: Does every exception have a review date? An exception without a review point tends to become invisible technical debt.
  • Evidence: Can reviewers see why a finding was accepted, dismissed, or classified as a false positive?
  • Reappearance: Does the workflow detect a finding that was fixed and later introduced again?
  • Retention: Are baseline versions and decision history retained long enough to support release review and compliance-ready vulnerability management?

A useful exception record contains the finding identifier, affected asset, severity, business impact, rationale, compensating controls, owner, approver, creation date, expiration date, remediation issue, and evidence link. Keep the record tied to a specific scope and version; a broad statement such as “known vulnerability” is not sufficient.

For AI vulnerability prioritization or triage automation, use AI to organize evidence and suggest priority, not to silently overwrite baseline decisions. A human should be able to inspect the inputs behind a suppression, risk change, or recommended exception.

Common mistakes

  • Baselining everything: This removes the immediate noise but can also hide serious findings. Baseline only reviewed, identifiable risk.
  • Using a branch baseline as a release baseline: A temporary development comparison can conceal a regression when it follows the branch into production.
  • Resetting after every scanner change: A reset destroys useful history. Compare configurations first and document the reason for any migration.
  • Ignoring disappeared findings: A missing result may indicate a fix, a scope change, an authentication failure, or a scanner error.
  • Matching only by line number: Refactoring can move code while leaving the underlying vulnerability unchanged. Use the strongest stable identity available.
  • Allowing permanent exceptions: Accepted risk should have an owner, rationale, and review or expiration date.
  • Separating the baseline from delivery workflow: If developers cannot see why a finding is known or what action is expected, they will treat the scanner as background noise.
  • Measuring success by finding count alone: A lower count may reflect reduced scope or changed rules. Review coverage, new findings, fixed findings, overdue exceptions, and reintroduced issues together.

When to revisit

Review the baseline before seasonal planning cycles, at a defined recurring interval, and whenever workflows or tools change. Also revisit it before a major release, after a significant dependency or framework update, when an internet-facing API changes, after a cloud or container architecture change, and following a security incident or material change in business risk.

Use this short review sequence:

  1. Confirm that the asset inventory and scan scope are current.
  2. Check for overdue exceptions and findings without owners.
  3. Compare new, resolved, changed, and reintroduced findings against the last approved release.
  4. Sample dismissed and accepted findings to confirm their evidence and rationale remain valid.
  5. Review scanner configuration, authentication, exclusions, and severity mappings.
  6. Decide whether to retain, amend, or replace the baseline, preserving the previous version and recording the reason.

When the workflow changes, update the baseline documentation alongside the pipeline configuration. When the application changes, compare the new scan with the last trusted release rather than with an arbitrary recent run. This keeps automated security scanning focused on meaningful regressions while preserving visibility into known risk.

For related pipeline decisions, see when a security scan should fail the build and the CI/CD security scanning checklist. Teams managing larger repositories can also review how to choose scan scope, ownership, and frequency for monorepos.

Related Topics

#security baselines#drift detection#CI/CD#regression testing#vulnerability scanning#release security
S

Scan Quest Editorial Team

Security Editorial Team

Senior editor and content strategist. Writing about technology, design, and the future of digital media. Follow along for deep dives into the industry's moving parts.