In snapd versions prior to 2.62, snapd failed to properly check the destination of symbolic links when extracting a snap. The snap format is a squashfs file-system image and so can contain symbolic links and other file types. Various file entries within the snap squashfs image (such as icons and desktop files etc) are directly read by snapd when it is extracted. An attacker who could convince a user to install a malicious snap which contained symbolic links at these paths could then cause snapd to write out the contents of the symbolic link destination into a world-readable directory. This in-turn could allow an unprivileged user to gain access to privileged information.

In snapd versions prior to 2.62, snapd failed to properly check the file type when extracting a snap. The snap format is a squashfs file-system image and so can contain files that are non-regular files (such as pipes or sockets etc). Various file entries within the snap squashfs image (such as icons etc) are directly read by snapd when it is extracted. An attacker who could convince a user to install a malicious snap which contained non-regular files at these paths could then cause snapd to block indefinitely trying to read from such files and cause a denial of service.

In snapd versions prior to 2.62, when using AppArmor for enforcement of sandbox permissions, snapd failed to restrict writes to the $HOME/bin path. In Ubuntu, when this path exists, it is automatically added to the users PATH. An attacker who could convince a user to install a malicious snap which used the 'home' plug could use this vulnerability to install arbitrary scripts into the users PATH which may then be run by the user outside of the expected snap sandbox and hence allow them to escape confinement.

When generating the systemd service units for the docker snap (and other similar snaps), snapd does not specify Delegate=yes - as a result systemd will move processes from the containers created and managed by these snaps into the cgroup of the main daemon within the snap itself when reloading system units. This may grant additional privileges to a container within the snap that were not originally intended.

The snapctl component within snapd allows a confined snap to interact with the snapd daemon to take certain privileged actions on behalf of the snap. It was found that snapctl did not properly parse command-line arguments, allowing an unprivileged user to trigger an authorised action on behalf of the snap that would normally require administrator privileges to perform. This could possibly allow an unprivileged user to perform a denial of service or similar.

Race condition in snap-confine's must_mkdir_and_open_with_perms()

Using the TIOCLINUX ioctl request, a malicious snap could inject contents into the input of the controlling terminal which could allow it to cause arbitrary commands to be executed outside of the snap sandbox after the snap exits. Graphical terminal emulators like xterm, gnome-terminal and others are not affected - this can only be exploited when snaps are run on a virtual console.

snapd 2.54.2 fails to perform sufficient validation of snap content interface and layout paths, resulting in the ability for snaps to inject arbitrary AppArmor policy rules via malformed content interface and layout declarations and hence escape strict snap confinement. Fixed in snapd versions 2.54.3+18.04, 2.54.3+20.04 and 2.54.3+21.10.1

snapd 2.54.2 did not properly validate the location of the snap-confine binary. A local attacker who can hardlink this binary to another location to cause snap-confine to execute other arbitrary binaries and hence gain privilege escalation. Fixed in snapd versions 2.54.3+18.04, 2.54.3+20.04 and 2.54.3+21.10.1

snapd 2.54.2 and earlier created ~/snap directories in user home directories without specifying owner-only permissions. This could allow a local attacker to read information that should have been private. Fixed in snapd versions 2.54.3+18.04, 2.54.3+20.04 and 2.54.3+21.10.1

A race condition existed in the snapd 2.54.2 snap-confine binary when preparing a private mount namespace for a snap. This could allow a local attacker to gain root privileges by bind-mounting their own contents inside the snap's private mount namespace and causing snap-confine to execute arbitrary code and hence gain privilege escalation. Fixed in snapd versions 2.54.3+18.04, 2.54.3+20.04 and 2.54.3+21.10.1

It was discovered that snapctl user-open allowed altering the $XDG_DATA_DIRS environment variable when calling the system xdg-open. OpenURL() in usersession/userd/launcher.go would alter $XDG_DATA_DIRS to append a path to a directory controlled by the calling snap. A malicious snap could exploit this to bypass intended access restrictions to control how the host system xdg-open script opens the URL and, for example, execute a script shipped with the snap without confinement. This issue did not affect Ubuntu Core systems. Fixed in snapd versions 2.45.1ubuntu0.2, 2.45.1+18.04.2 and 2.45.1+20.04.2.

cloud-init as managed by snapd on Ubuntu Core 16 and Ubuntu Core 18 devices was run without restrictions on every boot, which a physical attacker could exploit by crafting cloud-init user-data/meta-data via external media to perform arbitrary changes on the device to bypass intended security mechanisms such as full disk encryption. This issue did not affect traditional Ubuntu systems. Fixed in snapd version 2.45.2, revision 8539 and core version 2.45.2, revision 9659.

snap-confine as included in snapd before 2.39 did not guard against symlink races when performing the chdir() to the current working directory of the calling user, aka a "cwd restore permission bypass."

snap-confine in snapd before 2.38 incorrectly set the ownership of a snap application to the uid and gid of the first calling user. Consequently, that user had unintended access to a private /tmp directory.

Canonical snapd before version 2.37.1 incorrectly performed socket owner validation, allowing an attacker to run arbitrary commands as root. This issue affects: Canonical snapd versions prior to 2.37.1.

A vulnerability in the seccomp filters of Canonical snapd before version 2.37.4 allows a strict mode snap to insert characters into a terminal on a 64-bit host. The seccomp rules were generated to match 64-bit ioctl(2) commands on a 64-bit platform; however, the Linux kernel only uses the lower 32 bits to determine which ioctl(2) commands to run. This issue affects: Canonical snapd versions prior to 2.37.4.

In snapd 2.27 through 2.29.2 the 'snap logs' command could be made to call journalctl without match arguments and therefore allow unprivileged, unauthenticated users to bypass systemd-journald's access restrictions.