See on GitHub

CVE Binary Tool quick start / README

Build Status codecov Gitter On ReadTheDocs On PyPI Code style: black Imports: isort CII Best Practices Language grade: Python

The CVE Binary Tool is a free, open source tool to help you find known vulnerabilities in software, using data from the National Vulnerability Database (NVD) list of Common Vulnerabilities and Exposures (CVEs).

The tool has two main modes of operation:

  1. A binary scanner which helps you determine which packages may have been included as part of a piece of software. There are around 100 checkers which focus on common, vulnerable open source components such as openssl, libpng, libxml2 and expat.
  2. Tools for scanning known component lists in various formats, including .csv, several linux distribution package lists, language specific package scanners and several Software Bill of Materials (SBOM) formats.

It is intended to be used as part of your continuous integration system to enable regular vulnerability scanning and give you early warning of known issues in your supply chain.

For more details, see our documentation or this quickstart guide

Installing CVE Binary Tool

CVE Binary Tool can be installed using pip:

pip install cve-bin-tool

You can also do pip install --user -e . to install a local copy which is useful if you’re trying the latest code from the cve-bin-tool github or doing development. The Contributor Documentation covers how to set up for local development in more detail.

Finding known vulnerabilities using the binary scanner

To run the binary scanner on a directory or file:

cve-bin-tool <directory/file>

Note that this option will also use any language specific checkers to find known vulnerabilities in components.

Finding known vulnerabilities in a list of components

To scan a comma-delimited (CSV) or JSON file which lists dependencies and versions:

cve-bin-tool --input-file <filename>

Note that the --input-file option can also be used to add extra triage data like remarks, comments etc. while scanning a directory so that output will reflect this triage data and you can save time of re-triaging (Usage: cve-bin-tool -i=test.csv /path/to/scan). A VEX file (which may be created using the --vex command line option) can also be used as a triage file. A VEX file is detected if the file suffix is ‘.vex’.

Scanning an SBOM file for known vulnerabilities

To scan a software bill of materials file (SBOM):

cve-bin-tool  --sbom <sbom_filetype> --sbom-file <sbom_filename>

Valid SBOM types are SPDX, CycloneDX, and SWID.

Using the tool offline

Specifying the --offline option when running a scan ensures that cve-bin-tool doesn’t attempt to download the latest database files or to check for a newer version of the tool.

Note that you will need to obtain a copy of the vulnerability data before the tool can run in offline mode. The offline how-to guide contains more information on how to set up your database.

Output Options

The CVE Binary Tool provides console-based output by default. If you wish to provide another format, you can specify this and a filename on the command line using --format. The valid formats are CSV, JSON, console, HTML and PDF. The output filename can be specified using the --output-file flag.

The reported vulnerabilities can additionally be reported in the Vulnerability Exchange (VEX) format by specifying --vex command line option. The generated VEX file can then be used as an --input-file to support a triage process.

If you wish to use PDF support, you will need to install the reportlab library separately.

If you intend to use PDF support when you install cve-bin-tool you can specify it and report lab will be installed as part of the cve-bin-tool install:

pip install cve-bin-tool[PDF]

If you’ve already installed cve-bin-tool you can add reportlab after the fact using pip:

pip install --upgrade reportlab

Note that reportlab was taken out of the default cve-bin-tool install because it has a known CVE associated with it (CVE-2020-28463). The cve-bin-tool code uses the recommended mitigations to limit which resources added to PDFs, as well as additional input validation. This is a bit of a strange CVE because it describes core functionality of PDFs: external items, such as images, can be embedded in them, and thus anyone viewing a PDF could load an external image (similar to how viewing a web page can trigger external loads). There’s no inherent “fix” for that, only mitigations where users of the library must ensure only expected items are added to PDFs at the time of generation.

Since users may not want to have software installed with an open, unfixable CVE associated with it, we’ve opted to make PDF support only available to users who have installed the library themselves. Once the library is installed, the PDF report option will function.

Full option list

Usage: cve-bin-tool <directory/file to scan>

optional arguments:
  -h, --help            show this help message and exit
  -e, --exclude         exclude path while scanning
  -V, --version         show program's version number and exit
                        skips checking for a new version
  --detailed            display detailed report
                        skips checking xml files against schema
  --offline             operate in offline mode							

CVE Data Download:
  -n {json,api}, --nvd {json,api}
                        choose method for getting CVE lists from NVD
  -u {now,daily,never,latest}, --update {now,daily,never,latest}
                        update schedule for NVD database (default: daily)
  --nvd-api-key NVD_API_KEY
                    specify NVD API key (used to improve NVD rate limit)

  directory             directory to scan
  -i INPUT_FILE, --input-file INPUT_FILE
                        provide input filename
  -C CONFIG, --config CONFIG
                        provide config file
  -L PACKAGE_LIST, --package-list PACKAGE_LIST
                    provide package list
  --sbom {spdx,cyclonedx,swid}
                    specify type of software bill of materials (sbom)
                    (default: spdx)
  --sbom-file SBOM_FILE
                    provide sbom filename

  -q, --quiet           suppress output
  -l {debug,info,warning,error,critical}, --log {debug,info,warning,error,critical}
                        log level (default: info)
  -o OUTPUT_FILE, --output-file OUTPUT_FILE
                        provide output filename (default: output to stdout)  
  --html-theme HTML_THEME
                        provide custom theme directory for HTML Report
  -f {csv,json,console,html,pdf}, --format {csv,json,console,html,pdf}
                        update output format (default: console)
  -c CVSS, --cvss CVSS  minimum CVSS score (as integer in range 0 to 10) to
                        report (default: 0)
  -S {low,medium,high,critical}, --severity {low,medium,high,critical}
                        minimum CVE severity to report (default: low)
  --report              Produces a report even if there are no CVE for the
                        respective output format
  -A [<distro_name>-<distro_version_name>], --available-fix [<distro_name>-<distro_version_name>]
                        Lists available fixes of the package from Linux distribution
  -b [<distro_name>-<distro_version_name>], --backport-fix [<distro_name>-<distro_version_name>]
                        Lists backported fixes if available from Linux distribution
  --affected-versions   Lists versions of product affected by a given CVE (to facilitate upgrades)
  --vex VEX             Provide vulnerability exchange (vex) filename

Merge Report:
                        provide path for saving intermediate report 
  -t TAG, --tag TAG     provide a tag to differentiate between multiple intermediate reports
                        comma separated intermediate reports path for merging
  -F TAGS, --filter TAGS           
                        comma separated tags to filter out intermediate reports

  -s SKIPS, --skips SKIPS
                        comma-separated list of checkers to disable
  -r RUNS, --runs RUNS  comma-separated list of checkers to enable

Database Management:
  --export EXPORT       export database filename  
  --import IMPORT       import database filename  

   -x, --extract        autoextract compressed files
   CVE Binary Tool autoextracts all compressed files by default now

For further information about all of these options, please see the CVE Binary Tool user manual.

Note: For backward compatibility, we still support csv2cve command for producing CVEs from csv but we recommend using the --input-file command going forwards.

-L or --package-list option runs a CVE scan on installed packages listed in a package list. It takes a python package list (requirements.txt) or a package list of packages of systems that has dpkg, pacman or rpm package manager as an input for the scan. This option is much faster and detects more CVEs than the default method of scanning binaries.

You can get a package list of all installed packages in

  • a system using dpkg package manager by running dpkg-query -W -f '${binary:Package}\n' > pkg-list
  • a system using pacman package manager by running pacman -Qqe > pkg-list
  • a system using rpm package manager by running rpm -qa --queryformat '%{NAME}\n' > pkg-list

in the terminal and provide it as an input by running cve-bin-tool -L pkg-list for a full package scan.


You can use --config option to provide configuration file for the tool. You can still override options specified in config file with command line arguments. See our sample config files in the test/config

Using CVE Binary Tool in GitHub Actions

If you want to integrate cve-bin-tool as a part of your github action pipeline. You can checkout our example github action.

Binary checker list

The following checkers are available for finding components in binary files:

Available checkers

All the checkers can be found in the checkers directory, as can the instructions on how to add a new checker. Support for new checkers can be requested via GitHub issues.

Language Specific checkers

A number of checkers are available for finding vulnerable components in specific language packages.


The scanner examines the pom.xml file within a Java package archive to identify Java components. The package names and versions within the archive are used to search the database for vulnerabilities.

JAR, WAR and EAR archives are supported.


The scanner examines the package-lock.json file within a javascript application to identify components. The package names and versions are used to search the database for vulnerabilities.


The scanner examines the PKG-INFO and METADATA files for an installed Python package to extract the component name and version which are used to search the database for vulnerabilities.

The tool supports the scanning of the contents of any Wheel package files (indicated with a file extension of .whl) and egg package files (indicated with a file extension of .egg).

The --package-list option can be used with a Python dependencies file requirements.txt to find the vulnerabilities in the list of components.


This scanner does not attempt to exploit issues or examine the code in greater detail; it only looks for library signatures and version numbers. As such, it cannot tell if someone has backported fixes to a vulnerable version, and it will not work if library or version information was intentionally obfuscated.

This tool is meant to be used as a quick-to-run, easily-automatable check in a non-malicious environment so that developers can be made aware of old libraries with security issues that have been compiled into their binaries.

The tool does not guarantee that any vulnerabilities reported are actually present or exploitable, neither is it able to find all present vulnerabilities with a guarantee.

Users can add triage information to reports to mark issues as false positives, false negatives, indicate that the risk has been mitigated by configuration/usage changes, and so on.

Triage details can be re-used on other projects so, for example, triage on a Linux base image could be applied to multiple containers using that image.

For more information and usage of triage information with the tool kindly have a look here.

If you are using the binary scanner capabilities, be aware that we only have a limited number of binary checkers (see table above) so we can only detect those libraries. Contributions of new checkers are always welcome! You can also use an alternate way to detect components (for example, a bill of materials tool such as tern) and then use the resulting list as input to cve-bin-tool to get a more comprehensive vulnerability list.

The tool uses a vulnerability database in order to detect the present vulnerabilities, in case the database is not frequently updated (specially if the tool is used in offline mode), the tool would be unable to detect any newly discovered vulnerabilities. Hence it is highly advised to keep the database updated.


To use the auto-extractor, you may need the following utilities depending on the type of file you need to extract. The utilities below are required to run the full test suite on Linux:

  • file
  • strings
  • tar
  • unzip
  • rpm2cpio
  • cpio
  • ar
  • cabextract

Most of these are installed by default on many Linux systems, but cabextract and rpm2cpio in particular might need to be installed.

On windows systems, you may need:

  • ar
  • 7z
  • Expand
  • pdftotext

Windows has ar and Expand installed by default, but 7z in particular might need to be installed. If you want to run our test-suite or scan a zstd compressed file, We recommend installing this 7-zip-zstd fork of 7zip. We are currently using 7z for extracting jar, apk, msi, exe and rpm files.

If you get an error about building libraries when you try to install from pip, you may need to install the Windows build tools. The Windows build tools are available for free from

If you get an error while installing brotlipy on Windows, installing the compiler above should fix it.

pdftotext is required for running tests. (users of cve-bin-tool may not need it, developers likely will.) The best approach to install it on Windows involves using conda (click here for further instructions).

You can check our CI configuration to see what versions of python we’re explicitly testing.

Feedback & Contributions

Bugs and feature requests can be made via GitHub issues. Be aware that these issues are not private, so take care when providing output to make sure you are not disclosing security issues in other products.

Pull requests are also welcome via git.

  • New contributors should read the contributor guide to get started.
  • Folk who already have experience contributing to open source projects may not need the full guide but should still use the pull request checklist to make things easy for everyone.

Security Issues

Security issues with the tool itself can be reported to Intel’s security incident response team via

If in the course of using this tool you discover a security issue with someone else’s code, please disclose responsibly to the appropriate party.