PostgreSQL backup tool with Point-In-Time-Recovery and restore verification
TypeScript backupbackupsdatabasedatabase-backupdevopsdockergolangkubernetesmariadbmongodbmysqlpgpostgrespostgresqls3self-hostedsystem-administrationtoolsweb-ui
FEATURE (readme): Update readme security section
Rostislav Duginc7aa4d693cff
diff --git a/CLAUDE.md b/CLAUDE.mdindex 815b9d5e..fa1327d1 100644--- a/CLAUDE.md+++ b/CLAUDE.md@@ -59,4 +59,16 @@ After each change run linting and formatting depending on folder you are workingDon't write comments that explain previous behavior ("used to be X", "was renamed from Y", "kept for legacy callers"). Code shows the current state; history lives in git.-Don't preserve backward compatibility unless the user asks for it. No deprecation shims, no aliases, no fallbacks for the old shape. When planning a change that would break existing callers, schemas, configs, or APIs, call out the break explicitly in the plan. If the user approves it, delete the old code outright — do not leave a transition layer behind.\ No newline at end of file+Don't preserve backward compatibility unless the user asks for it. No deprecation shims, no aliases, no fallbacks for the old shape. When planning a change that would break existing callers, schemas, configs, or APIs, call out the break explicitly in the plan. If the user approves it, delete the old code outright — do not leave a transition layer behind.++### Security++Databasus handles sensitive data, so security is a layered defence. CodeQL, CodeRabbit, Codex Security, Trivy, Dependabot, gitleaks and semgrep all run on every PR. When working in the repo:++- Don't disable or weaken security checks to make a build pass. For genuine false positives, suppress explicitly with a documented reason (e.g. `.trivyignore`) and explain why in the PR.+- Pin every new GitHub Action to a full commit SHA with a `# vX.Y.Z` tag comment. No floating tags like `@v4` or `@main`.+- Workflows default to top-level `permissions: contents: read`; elevate per-job only when justified.+- Keep Dockerfiles free of secrets, floating base-image tags and unjustified root. If root-at-start is required (PUID/PGID remap, volume chown, initdb), drop privileges with `gosu` before `exec`-ing the app.+- Never log secrets, tokens or credentials. Redact at the logger layer, not at call sites.++The README's `🛡️ Security & reliability engineering` section is the public-facing version of these practices — keep both consistent if substance changes.\ No newline at end of filediff --git a/README.md b/README.mdindex 38990d82..fea62aad 100644--- a/README.md+++ b/README.md@@ -249,6 +249,22 @@ After installation, it is also recommended to <a href="https://databasus.com/faq---+## 🛡️ Security & reliability engineering++Databasus works with sensitive data, so preventing vulnerabilities, unauthorised access and data leaks is a primary concern. We invest in this on both sides of the system: in the code itself (permission checks, encryption, careful handling of secrets) and in the infrastructure around it (dependency analysis, CVE response, DevSecOps best practices). The pipeline below runs automatically on every commit and PR - no single layer is enough on its own, but together they catch most things before a release goes out.++For static analysis we combine several independent passes. CodeQL scans the full codebase for security issues. CodeRabbit reviews every PR and runs gitleaks for secret scanning and semgrep for security rules inline. Dockerfiles and CI workflows get extra rules of their own (pinned action references, least-privilege permissions, suspicious base images), so insecure patterns are flagged before they ever merge. On top of these per-PR checks, Codex Security from OpenAI runs regular, deeper audits of the whole codebase. It's a separate program that catches architectural and cross-cutting issues narrow PR-time scans can miss.++On the dependency side, Dependabot watches all of our dependencies against the GitHub Advisory Database and surfaces CVEs within minutes of publication. Updates run through a cooldown so newly-published versions get a chance to mature before we adopt them. This is a deliberate defence against compromised-package incidents like supply-chain attack. The Dependency Review Action blocks any PR that introduces a new HIGH or CRITICAL CVE outright.++Container images are scanned with Trivy on every build. A separate Trivy pass on the Dockerfile catches misconfigurations before they make it into an image. All GitHub Actions are pinned to full commit SHAs rather than floating tags like `@v4` or `@main`, which have been an active attack vector in 2025. Workflows default to least-privilege permissions and only elevate per-job when genuinely needed.++Critical paths are covered by both unit and integration tests, run against real database containers for every supported engine and major version. Restore is the path that matters most for a backup tool, so we test it explicitly: every PR runs full backup-then-restore cycles against those same real containers, verifying that backups can actually be restored end-to-end, not just written successfully. The rest of the CI/CD pipeline runs lint, type-check, the full test suite, image smoke tests and multi-architecture builds on every PR. A release only ships if all of it passes.++Found a vulnerability? Report it via the GitHub Security tab. See [SECURITY.md](https://github.com/databasus/databasus?tab=security-ov-file#readme). Security reports are the highest-priority work queue. For runtime application security (AES-256-GCM at rest, zero-trust storage, encrypted secrets, read-only DB user by default) see [Enterprise-grade security](#-enterprise-grade-security) in the Features section above.++---+## 📝 LicenseThis project is licensed under the Apache 2.0 License - see the [LICENSE](LICENSE) file for details@@ -284,19 +300,13 @@ AI is not used for:- code without line-by-line verification by a human- code without tests-The project has:--- solid test coverage (both unit and integration tests)-- CI/CD pipeline automation with tests and linting to ensure code quality-- verification by experienced developers with experience in large and secure projects-So AI is just an assistant and a tool for developers to increase productivity and ensure code quality. The work is done by developers.Moreover, it's important to note that we do not differentiate between bad human code and AI vibe code. There are strict requirements for any code to be merged to keep the codebase maintainable.Even if code is written manually by a human, it's not guaranteed to be merged. Vibe code is not allowed at all and all such PRs are rejected by default (see [contributing guide](https://databasus.com/contribute)).-We also draw attention to fast issue resolution and security [vulnerability reporting](https://github.com/databasus/databasus?tab=security-ov-file#readme).+The engineering safeguards behind these rules (CI, static analysis, dependency scanning, test coverage and vulnerability response) are documented in [Security & reliability engineering](#️-security--reliability-engineering) above.### You have a cloud version — are you truly open source?