Debian Security Team

Debian Security Tracker


Everything in the Debian Security Tracker is publicly available, as in "Debian doesn't hide problems" available.

The best thing about our tracking system is that it is very basic. There is no overhead of web-based ticket/issue trackers, it's just a Subversion (SVN) repository and some text files that we collaboratively edit and then some scripts to parse these files and generate useful reports available online. Everything is designed to be very simple to use, transparent and easy to see what other people are working on so you can work on other things.

Gentle Introduction

The following will give you a basic walkthrough of how the files are structured, and how we do our work while tracking issues.

The best way to understand is to check out our repository from Subversion so you have the files on your computer and can follow along at home. To do this you just need to do the following:

svn co svn+ssh://<alioth user name>

This will check out the working repository (given that you already have an Alioth account and public key authentication already set up. After successful downloading, you will have a new directory called secure-testing. Inside this directory are a number of subdirectories. The data directory is where we do most of our work.

Note that the name of the Subversion repository is historical; the tracker is not specially related to testing-security, but for Debian security at large.

If you don't have an Alioth account, you can create one. You can then join the team by clicking the Request to join link.

If you don't need write access, you can of course check out our files without an Alioth account as well:

svn co svn://

If you are a Git fan, you can also use git-svn. Once you have the git-svn package installed, you can clone the Subversion repository into your own local Git repository with:

git svn clone svn+ssh://<alioth user name>

Note that this will take a very long time (expect over two hours) since every commit from the very beginning (over 12,000 at this point) is checked out individually and merged into your Git repository.

Subversion and git-svn Crash Course

The following table lists the most common/useful commands for working with the secure-testing repository:

subversion git-svn action
svn update git svn rebase sync your local repo from remote secure-testing repo
svn commit git svn dcommit commit your changes to the remote secure-testing repo (note that git commit -a only updates your local repo)
svn diff git diff compare your local repo to remote secure-testing repo

The CVE list (CVE/list)

Automatic Issue Updates

Twice a day a cron job runs that pulls down the latest full CVE lists from MITRE, automatically checks that in into data/CVE/list, and also syncs that file with other lists like data/DSA/list and data/DTSA/list.

These automatic commits as well as all Subversion commits are notified via either the secure-testing-commits mailing list, or via the KGB IRC bot in the #debian-security channel on the OFTC IRC network. For example, the bot could say in the channel:

17:14 < KGB-0> sectracker r21191 data/CVE/list * automatic update

Most of our work consists of taking new issues that MITRE releases and processing them so that the tracking data is correct. Read on for an explanation of how we do this.

Processing TODO entries

The MITRE update typically manifests in new CVE entries. So what we do is update our Subversion repository and then edit data/CVE/list and look for new TODO entries. These will often be in blocks of 10-50 or so, depending on how many new issues have been assigned by MITRE.

Processing TODO entries means checking if the problem affects Debian and if so which packages, as well as evaluate their severity. This information is based on research and not just in the CVE description in order to prevent integrating false positives or incorrect data in the security tracker. For example, if the CVE id says that something is vulnerable prior to version X, you need to check if that is the case as well as for the information given on distribution specific issues. Always make sure you understand the issue and are able to verify that the information is correct.

Thus, a proper research should include reading the code, finding fixes/commits in the upstream repository, or even writing patches yourself if you have the time and skills to do that. If you can't assure that, please add a TODO entry reflecting what is missing from your research. Check the section NOTE and TODO entries for more details.

If you are aware of an error in some CVE description, please write to the oss-security mailing list, with a carbon copy (cc) to


Processing entries is done by first seeing if the issue is related to any software packaged in Debian. If it isn't a package in Debian and has no ITP/RFP then you make a note of that in the file using a NOT-FOR-US: tag. Third-party modules not yet packaged for Debian are also tagged as NFU; even if their parent software is packaged for Debian. The module names should be mentioned in the NFU note in order to make issues apparent if that module should ever receive a proper package. Another case are meta packages that only provide a downloader (e.g., flashplugin-nonfree). There is no way to mark such packages as we have no influence on the version and technically the code is not present in Debian.


CVE-2005-3018 (Apple Safari allows remote attackers to cause a denial of service ...)
        NOT-FOR-US: Safari

Before marking a package NFU, the following should be done:

If there is any doubt, add a NOTE with your findings and/or ask others to double check using TODO (see NOTE and TODO entries below).

There is a tool that helps with sorting out all the NOT-FOR-US issues: bin/check-new-issues -h. For the search functions in check-new-issues to work, you need to have unstable in your sources.list and have done apt-get update and apt-file update. Having libterm-readline-gnu-perl installed helps, too. If you are not running unstable, you can search at or set up an unstable chroot.

Packages in the archive

If the vulnerability refers to a package in the Debian archive (except for experimental, see later), look to see if the package is affected or not (sometimes newer versions that have the fixes have already been uploaded).

If the version has been fixed already, note the package name and the Debian version that fixes it and assign a severity level to it, for example:

CVE-2005-2596 (User.php in Gallery, as used in Postnuke, allows users with any Admin ...)
        - gallery 1.5-2 (medium)

Even if the CVE description mentions it is fixed as of a particular version, double-check the Debian package yourself (because sometimes the CVE descriptions or information from databases like Secunia is incorrect).

If it hasn't been fixed, we determine if there has been a bug filed about the issue, and if not, file one and then note it in the list (again with a severity level):

CVE-2005-3054 (fopen_wrappers.c in PHP 4.4.0, and possibly other versions, does not ...)
        - php4 <unfixed> (bug #353585; medium)
        - php5 <unfixed> (bug #353585; medium)

Bug numbers can be added as in the example above. To avoid duplicate bugs, bug filed can be added instead of bug #123456 when the bug report has been sent but the bug number is not yet known (however, it is more desirable to file the bug, wait for the BTS to assign a number, then update the entry in the CVE list so that complete information is always available in the tracker). The bug number is important because it makes it clear that the maintainer has been contacted about the problem, and that they are aware of their responsibility to work swiftly toward a fix.

Since CVEs often drop in bulk, submission of multiple CVEs in a single bug report is permissible and encouraged. However, some maintainers have indicated a preference for only one issue per bug report. The following is a list of packages for which each CVE should be reported separately:

A special exception is made for kernel related issues. The kernel-sec group will take care of them. It is not necessary to file bugs in the BTS for kernel security issues, it only causes overhead.

If you want to report a bug, bin/report-vuln might be helpful in creating the bug report.

If a vulnerability does not affect Debian, e.g., because the vulnerable code is not contained, it is marked as :

CVE-2004-2628 (Multiple directory traversal vulnerabilities in thttpd 2.07 beta 0.4, ...)
        - thttpd <not-affected> (Windows-specific vulnerabilities)

<not-affected> is also used if a vulnerability was fixed before a package was uploaded into the Debian archive.

Undetermined Tags

If you don't have time to fully research an issue, but it is abundantly clear (via CVE text or other announcement) that the issue affects a particular package or set of packages, the tag can be used. This has the advantage of entering the issue earlier in the output of debsecan and on the PTS pages, which is useful for the small set of proactive maintainers paying attention to these information sources. Getting the maintainer involved hopefully prompts faster fixes. This also allows enables tracking of multiple packages, some of which may already be fixed.

<undetermined> can also be used when there simply is not enough information disclosed in the existing known references about the issue. Essentially, <undetermined> indicates that someone needs to come back and revisit the issue. An example undetermined entry is:

CVE-2011-2351 (Use-after-free vulnerability in Google Chrome before 12.0.742.112 ...)
        - chromium-browser 12.0.742.112~r90304-1
        - webkit <undetermined>
        NOTE: webkit commit #123456

The list of all of currently undetermined issues is aggregated by the tracker. This is a good place for new contributors to get started since these are issues that can be pruned quickly for new information that may not have been known during the initial disclosure, and thus marked for further work or closed with a version number. Please add notes if you do change an undetermined issue to unfixed (unless you're also fixing the issue in the process, which is of course the ideal way to help/contribute).

Packages in Experimental only

There are some packages that only exists in experimental. In that case, place the distribution tag experimental. For example:

CVE-2013-1067 (Apport 2.12.5 and earlier uses weak permissions for core dump files ...)
        [experimental] - apport 2.12.6-1 (bug #727661)

If the package is in unstable and in experimental, focus on unstable (we are not tracking fixes in experimental). A note about the situation in experimental is appreciated though. For example:

CVE-2014-8564 (The _gnutls_ecc_ansi_x963_export function in gnutls_ecc.c in GnuTLS ...)
        - gnutls28 <unfixed> (bug #769154)
        NOTE: in experimental fixed in 3.3.10-1

Issues in ITP and/or RFP packages

If an issue is discovered in a package that has an RFP or ITP already filed, then that is also noted in order to track the problem, and made sure it is resolved before the package enters the archive. These issues are marked with the tag. Note this includes both ITPs and RFPs since (from a security tracking standpoint) there is no advantage in tracking them in separate ways. An example entry for an ITP/RFP package is:

CVE-2004-2525 (Cross-site scripting (XSS) vulnerability in compat.php in Serendipity ...)
        - serendipity <itp> (bug #312413)

Reserved entries

Several security problems have coordinated dates of public disclosure, i.e., a CVE identifier has been assigned to a problem, but it's not public yet. Also, several vendors have a pool of CVE ids they can assign to problems that are detected in their products. Such entries are marked as RESERVED in the tracker:


Rejected entries

Sometimes there are CVE assignments that later turn out to be duplicates, mistakes or non-issues. These items are reverted and turned into REJECTED entries:


Removed packages

Sometimes there are cases, where a vulnerability hasn't been fixed with a code change, but simply by deciding that a package is broken so severely that it needs to be removed from the archive entirely. This is tracked with the <removed> tag:

CVE-2005-1435 (Open WebMail (OWM) before 2.51 20050430 allows remote authenticated ...)
        - openwebmail <removed>

Also note that it is sufficient to mark a package as removed in unstable. The tracker is aware of which package is present in which distribution and marks other distributions that still contain the package automatically as unfixed. For example, if libxml is in oldstable, but not stable or unstable, then:

    - libxml <removed>

will track oldstable as affected, but stable and unstable as not-affected.

Once a package has been completely removed from all currently supported Debian releases, it should be tracked in the data/packages/removed-packages file. This file lists all packages (one source package per line) that were at one time in a Debian release, but no longer exist in any supported version. Additions to this file can be used to address failing consistency checks after a new release.

end-of-life packages

In rare cases (i.e., webbrowsers) security support for packages needs to be stopped before the end of the regular security maintenance life cycle.

Packages which are not anymore supported by the security team in a (old-)stable release are marked with the end-of-life tag:

CVE-2011-3973 (cavsdec.c in libavcodec in FFmpeg before 0.7.4 and 0.8.x before 0.8.3 ...)
        - libav 4:0.7.1-7 (bug #641478)
        - ffmpeg <removed>
        - ffmpeg-debian <end-of-life>

NOTE and TODO entries

There are many instances where more work has to be done to determine if something is affected, and you might not be able to do this at the time. These entries can have their TODO line changed to something descriptive so that it is clear what remains to be done. For example:

CVE-2005-3990 (Directory traversal vulnerability in FastJar 0.93 allows remote ...)
        TODO: check, whether fastjar from the gcc source packages is affected

If you are not sure about some decision (e.g., which package is affected) or triaging (e.g., bug severity) you can leave a TODO note for reviewing, explaining which aspect have to be reviewed. For example:

CVE-2013-7295 (Tor before, when OpenSSL 1.x is used in ...)
        - tor (low)
        [wheezy] - tor <no-dsa> (Minor issue)
        TODO: review, severity. The exploitation scenario is too complicated.

It is also useful to add information to issues as you find it, so that when others go to look at an issue and want to know why you marked it as you did, or need a reference, it will be there. The more information left, the better. For example, the following entry lets you know that CVE-2005-3258 doesn't affect the squid that we have because the issue was introduced in a patch that was never applied to the Debian package:

CVE-2005-3258 (The rfc1738_do_escape function in ftp.c for Squid 2.5 STABLE11 and ...)
        - squid <not-affected> (bug #334882; medium)
        NOTE: Bug was introduced in a patch to squid-2.5.STABLE10,
        NOTE: this patch was never applied to the Debian package.

Severity levels

These levels are mostly used to prioritize the order in which security problems are resolved. Anyway, we have a rough overview on how you should assess these levels.

unimportant: This problem does not affect the Debian binary package, e.g., a vulnerable source file, which is not built, a vulnerable file in doc/foo/examples/, PHP Safe mode bugs, path disclosure (doesn't matter on Debian). All "non-issues in practice" fall also into this category, like issues only "exploitable" if the code in question is setuid root, exploits which only work if someone already has administrative privileges or similar. This severity is also used for vulnerabilities in packages which are not covered by security support.

low : A security problem, which has only mild security implications (local DoS, /tmp file races and so on).

medium : For anything which permits code execution after user interaction. Local privilege escalation vulnerabilities are in this category as well, or remote privilege escalation if it's constrained to the application (i.e., no shell access to the underlying system, such as simple cross-site scripting). Most remote DoS vulnerabilities fall into this category, too.

high : A typical, exploitable security problem, which you'll really like to fix or at least implement a workaround. This could be because the vulnerable code is very broadly used, because an exploit is in the wild or because the attack vector is very wide. Should be put into that category anything that permits an attacker to execute arbitrary code on the vulnerable system (with or without root privileges) and high-impact denial-of-service bugs (for instance, an IPv4 forwarding path vulnerability which requires only very few packets to exploit). Significant defects in security software can be rated "high" as well (for instance, a vulnerability in a piece of cryptographic software which flags forged digital signatures as genuine).

Certain packages may get higher or lower rating than usual, based on their importance.

Vulnerabilities without an assigned CVE id

If you learn of a vulnerability to which no CVE id has been assigned yet, you can request one. To request a CVE for public issues, you can write to the moderated oss-security list. In the meantime, you can add an entry of the form

CVE-2009-XXXX [optipng array overflow]
        - optipng (low)

It is desirable to include references which uniquely identify the issue, such as a permanent link to an entry in the upstream bug tracker, or a bug in the Debian BTS. If the issue is likely present in unstable, a bug should be filed to help the maintainer to track it.

Lack of CVE entries should not block advisory publication which are otherwise ready, but we should strive to release fully cross-referenced advisories nevertheless.

CVE pool from Debian

Debian can only assign CVE numbers from its own pool for issues which are not public. To request a CVE from the Debian pool, write to and include a description which follows CVE conventions.

Distribution tags

Our data is primarily targeted at sid, as we track the version that a certain issue was fixed in sid. The Security Tracker web site (see below) derives information about the applicability of a vulnerability to stable and oldstable from the list of DSAs issued by the security team and the fact that a source package is part of a release. Distribution tags can be used to denote information about a vulnerability for the version of a package in a specific release. An example:

CVE-2005-3974 (Drupal 4.5.0 through 4.5.5 and 4.6.0 through 4.6.3, when running on ...)
        - drupal 4.5.6-1 (low)
        [sarge] - drupal <not-affected> (Only vulnerable if running PHP 5)

Drupal has been fixed since 4.5.6, however Drupal from Sarge still isn't vulnerable as the vulnerability is only effective when run under PHP 5, which isn't part of Sarge.

When a vulnerability is fixed in (oldstable-)proposed-updates, it is added to next-(oldstable-)point-update.txt and only added to CVE/list after the point release (during which the no-dsa entry is removed).

Generated Reports

All of this tracking information gets automatically parsed and compared against madison (a program which inspects a local Debian package archive and displays the versions of the given packages found in each suite) to determine what has been fixed and what is still waiting, this results in this website:

It incorporates package lists and parses distribution lists and can thus be used to:

For every security problem it displays:

The DSA list (DSA/list)

We maintain a list of all DSA advisories issued by the stable security team. This information is used to derive information about the state of security problems for the stable and oldstable distribution. An entry for a DSA looks like this:

[21 Nov 2005] DSA-903-1 unzip - race condition
        [woody] - unzip 5.50-1woody4
        [sarge] - unzip 5.52-1sarge2
        NOTE: fixed in testing at time of DSA

The first line tracks the date when a DSA was issued, the DSA identifier, the affected source package, and the type of vulnerability. The second line performs a cross-reference to the entry in CVE/list that maintains the state of the vulnerability in sid. Every entry that is added like this to DSA/list is parsed by a script and automatically added to CVE/list. The next lines contain the fixes for stable and optionally oldstable, addressed with distribution tags. You may add NOTE: entries freely, we use a NOTE entry for statistical purposes that tracks when a fix has reached testing relative to the time when it hit stable.

There is no need to add anything to CVE/list for a DSA, the DSA cross-reference will be added automatically by the cron job. However, you do need to add [lenny] or [squeeze] entries to CVE/list when there is a no-dsa or not-affected condition.

Checking in your changes

After thoroughly researching each issue (as described above) and editing the relevant files, commit your changes. Peer review is (hopefully) done via the mailing list and IRC notifications (see Automatic issue updates above). However, changes to the tracker website itself (e.g., the files in lib/* and bin/ should be vetted and approved before being committed. The preferred way to do this is to send a patch to the mailing list.

Commits are checked for syntax errors before they are actually committed, and you'll receive an error and your commit is aborted if it is in error. To check your changes yourself beforehand, use make check-syntax from the root of the SVN directory.

Following up on security issues

By simply loading this page and doing a little gardening of the different issues, many things can be done. One thing is that you can read all the bug reports of each issue and see if new information has been added to the end that might provide updated or changed information (such as if an issue has been closed, or a version of the package has been uploaded that contains the fix). It is also useful to follow-up on the issues to prod the maintainer to deal with the issue, which they may have forgotten about.

Tracking of security bugs in the BTS and linking them to a user tag by CVE

There's an automated tagging of security-related bugs to CVE IDs through the user tag security for the user

All bugs added to the tracker are automatically tagged. You can use the search here to find all bugs not yet present in the tracker.

All bug numbers added to the tracker are automatically associated with the relevant user tag.

If you checked an issue which doesn't need to be added to the tracker (e.g., because it's not security-relevant or otherwise bogus) you can either remove the security tag from the bugs or send a mail to with the following content:

usertag $BUGNUM + tracked

Contributing with the security tracker code


Setting up a local testing instance

It is possible to set up an instance of the security tracker in your own machine for testing purposes. The following packages are needed:


The following commands build the databases for stable and run a python local server in port 10605:

make update-stable
make serve

The website is now available as