Inventory and SBOM
Know everything inside your images, including what does not come from a package: copied binaries, libraries built into other binaries and pasted code. Export it as an SBOM in CycloneDX or SPDX.
scancontainer reads every file of a container image and tells you what it is, where it came from and what to fix. Here is what that covers.
Know everything inside your images, including what does not come from a package: copied binaries, libraries built into other binaries and pasted code. Export it as an SBOM in CycloneDX or SPDX.
Each file is compared with the official signed package from its distribution, not with the checksums the image carries. You get a verdict: intact, modified or with unexplained content.
Miners, webshells, reverse shells, rootkits, droppers and known malware families, found without running anything. Hidden and encoded code is unpacked and read too.
Less list, more proof. For each vulnerability: does it affect this version, does it really apply, can it be reached, is it being exploited, and what do I update.
The better base image and the Dockerfile line to change. Root user, exposed secrets, build tools left behind and weak binary hardening, checked against CIS and NSA/CISA guidance.
Hundreds of kinds of credentials, found in files, environment variables, build history and deleted layers. Validated offline and always shown masked.
The licence of every component, normalised to SPDX, and a calendar of the systems and runtimes that are losing support.
One summary for managers and one detailed report for engineers, both reproducible. Exports in CycloneDX, SPDX, OpenVEX, CSAF, SARIF, JSON, HTML and PDF.
In detail
Coverage
What makes it different
Each file in the image is classified and compared with its official source. What cannot be read is reported as a gap, never passed as clean.
The file, the line, the reason and where the data came from. You can audit a finding and argue with it.
For each vulnerability we say whether it applies to your image and whether it can be reached, so the report starts with what matters.
Binaries copied by hand, libraries built into other binaries and pasted code. This is where package-only scanners go blind.
We read the image without starting it. Scanning a hostile image does not execute it.
No binary, code or secret is sent to an outside service. Results do not depend on an AI model, so the same image always gives the same report.
On the way
Coming soon
An agent on each Kubernetes node that only lets run what the image brought with it.
Coming soon
Start the image in an isolated microVM with no real network and see what it tries to do.
Coming soon
Attack paths from the internet to your data across AWS, GCP and Azure.
Coming soon
Models inspected without loading them and agent configuration reviewed.
Coming soon
The same product inside your own infrastructure, with no internet connection.
Same team, same short reports
scancodebase checks the code that goes into the image: security flaws, dependencies and secrets.