Where Secrets Really End Up in Docker Images: Testing 8 Common Methods
A comprehensive technical investigation examined exactly where secrets end up when building Docker images using eight different techniques. The tests used Ubuntu 24.04.1 LTS, Docker Engine 29.1.3, and BuildKit v0.26.2, with a test secret string SUPER_SECRET_KOTY_HULIGANY_2026 placed in various ways.
The first method, COPY, adds the secret file directly into an image layer. Even after running docker history, the file remains fully recoverable by extracting layers from a saved tar archive. Adding a subsequent RUN rm command only marks the file as deleted in a new layer; the original data stays intact in the previous layer.
Using ENV stores the secret value in the image configuration JSON under config.Env and also appears in build history. The ARG instruction keeps the value in history.created_by fields even though it does not appear in the final container environment.
The dedicated BuildKit Secrets feature, invoked with --mount=type=secret and --secret build flag, mounts the secret only temporarily during the RUN step. After the build completes, the value is absent from all layers, configuration JSON files, and history entries.
Multi-stage builds also proved effective when the secret is created and used only in an intermediate stage and never copied into the final image. No trace of the secret appears in the published artifact.
Combining echo and rm inside a single RUN instruction still leaves the secret string visible in build history and created_by metadata. Overwriting a copied file with a new value similarly fails because the original layer remains unchanged.
The experiments confirm that only two approaches — BuildKit Secrets and properly isolated multi-stage builds — avoid embedding secrets in the final Docker image. All other common patterns leave extractable data that can be recovered without ever starting a container.
Related articles
CrowdSec Confirms Theft of Source Code from Roughly 300 GitHub Repositories via TanStack Supply Chain Attack
French cybersecurity firm CrowdSec has confirmed that attackers stole source code from approximately 300 GitHub repositories, including around 170 private ones. The breach occurred in May 2026 through a compromised TanStack component that exfiltrated an API key with read access to the private codebase. The stolen material included code for the company's SaaS console, AWS procedures, connectors, and automation tools, while the remaining repositories contained already-public open source code. No customer data, passwords, organization details, tokens, or other secrets were included in the leak, and all potentially affected credentials were immediately rotated. CrowdSec stated that the code is tightly integrated with internal systems and has largely changed over the past four months, reducing its usefulness outside the company's environment. The SaaS service code undergoes regular audits, and the company sees no immediate threat from the exposure while the investigation continues.
Dependency Confusion Attacks Let Attackers Hijack Internal Library Names in Corporate Builds
A widespread supply chain risk allows attackers to publish packages with internal company names on public registries such as npm and PyPI, causing build systems to pull malicious versions instead of internal ones. The attack works because package managers treat multiple registries as a single list and select the highest version number, with no inherent priority for internal sources. Researcher Alex Birsan demonstrated the technique in February 2021 by registering names harvested from open repositories and error messages, successfully injecting packages into builds at Microsoft, Apple, PayPal, Shopify, Netflix, Tesla and Uber. The malicious code executes during installation because setup scripts and lifecycle hooks run with the privileges of the build agent, exposing environment variables, tokens and internal network access. Mitigation requires a single internal proxy repository that never mixes public responses for internal package names, scoped namespaces bound to private registries, lock files with content hashes, and disabling install scripts where possible. The technique remains effective against any organization that lists both internal and public registries in its build configuration.
NEOMSA ESB Release Strengthens Supply Chain Security Through SBOM and Dependency Hardening
Neoflex has released a new version of its NEOMSA ESB integration platform with a primary focus on cleaning up the software bill of materials and eliminating critical and high-severity vulnerabilities. The team automated SBOM generation using CycloneDX, ran SCA scans with Grype and OWASP Dependency-Check, and performed SAST and secret scanning across all build pipelines. Instead of blindly updating to the latest versions, engineers applied minimal fixed versions for each advisory while handling complex cases involving transitive dependencies, locked files, and deprecated build tools. The effort reduced the total package count from 5,872 to 1,683 after migrating the frontend build to Vite in Camel Karavan 4.18. Remaining medium and low findings were tracked in DefectDojo with clear remediation timelines. The changes deliver measurable risk reduction for on-premise deployments in critical infrastructure and financial organizations.
CodeScoring Launches CodeScoring.Save Artifact Repository for Secure Enterprise Development
CodeScoring has introduced its own artifact storage solution called CodeScoring.Save, designed to handle packages, libraries, container images, and other software components used in development. The product targets corporate users of any size seeking a predictable and resilient repository that integrates security checks directly into storage and distribution workflows. Built in Go for modern Kubernetes environments, Save supports multiple package formats including Maven, npm, NuGet, PyPI, Go Modules, Docker/OCI, DEB, and RPM while providing proxy access to external repositories. It features role-based access, auditing, independent scaling of compute and storage layers, and native integration with CodeScoring.OSA to surface vulnerability data inside the repository itself. The company positions Save as a standalone local deployment option that can operate independently or alongside its existing OSA Proxy module to block malicious components at the repository level. Future plans include support for AI models as artifacts, starting with storage and distribution for ecosystems such as Hugging Face, along with certification for Russian secure development requirements.