Vulnerabilities

Kata Containers is an open source project focusing on a standard implementation of lightweight Virtual Machines (VMs) that perform like containers. Versions prior to 3.31.0 ship with a default configuration that allows pod creators to inject arbitrary command-line arguments into the virtiofsd process through the `io.katacontainers.config.hypervisor.virtio_fs_extra_args` pod annotation. By injecting `-o source=/` along with `--no-announce-submounts` and `--sandbox=none`, an attacker can override the virtiofsd shared directory to serve the entire host root filesystem into the guest VM. Combined with the `kernel_params` annotation (also enabled by default) to activate the agent debug console, the attacker can mount the host filesystem from inside the VM and read or write any file on the host, including /etc/shadow. Version 3.31.0 patches the issue.

Incus is a system container and virtual machine manager. Prior to version 7.0.0, backup.GetInfo() trusts the inline backup/index.yaml config when present and only falls back to parsing the legacy backup/container/backup.yaml file if result.Config == nil. As a result, an archive can carry a valid inline config that passes the initial import preflight while also carrying a malformed legacy backup/container/backup.yaml file that is reparsed later from the restored file system. ParseConfigYamlFile() accepts YAML documents with no container section, and multiple downstream consumers then dereference. Container without checking for nil. Confirmed examples in the instance restore and import flow include backup.UpdateInstanceConfig() and internalImportFromBackup(). An authenticated user with permission to import instance backups may be able to crash the Incus daemon with a crafted backup archive whose inline backup/index.yaml is valid but whose extracted legacy backup.yaml omits container. The crash occurs in the restore path after archive extraction has begun. This issue has been patched in version 7.0.0.

Incus is a system container and virtual machine manager. In versions before 7.0.0, missing validation logic in the storage volume import logic allows an authenticated user with access to the storage volume feature to cause the Incus daemon to crash. The backup restore subsystem contains an out-of-bounds panic vulnerability caused by an invalid bounds check when indexing snapshot metadata arrays, and the same flawed pattern also appears in the migration path. When iterating through physical snapshots provided in a backup archive, the loop uses the index to look up corresponding metadata in the parsed `Config.Snapshots` and `Config.VolumeSnapshots` slices. The guard condition `len(slice) >= i-1` is incorrect because it can still evaluate to true when the subsequent slice[i] access is out of bounds. An attacker can submit a backup archive that contains physical snapshot directories while supplying a tampered `index.yaml` with an empty or truncated snapshot metadata array, causing the daemon to index beyond the end of the metadata slice and crash. Repeated use of this issue can be used to keep Incus offline, causing a denial of service. This issue is fixed in version 7.0.0.

lxc is a Linux container runtime. In the setuid helper lxc-user-nic, the delete path contains a logic flaw in the find_line() function that allows an unprivileged user to delete OVS-attached network interfaces belonging to other users. When lxc-user-nic delete scans its NIC database to authorize a deletion request, the interface name comparison can set the authorization flag based on a name match alone, even when the ownership, type, and link fields in that database entry belong to a different user. The vulnerable check sits after the goto next label handling, meaning it is reachable on lines where earlier ownership checks failed or were skipped. Because nothing downstream of this authorization signal re-verifies that the matched database line actually belongs to the caller, an unprivileged attacker with a valid lxc-usernet policy entry can trigger deletion of another user's OVS port on the same bridge. This is limited to multi-tenant environments using lxc-user-nic with OpenVSwitch bridges. The impact is denial of service - one tenant can repeatedly disconnect networking from containers run by another tenant on shared infrastructure. This is patched in version 7.0.0.

Incus is an open source container and virtual machine manager. In versions prior to 7.0.0, the image import flow issues an outbound HEAD request to a user-supplied URL before validating the request against project restrictions such as restricted.images.servers. The imgPostURLInfo function constructs and sends a HEAD request directly from the attacker-supplied source URL to resolve image metadata, and this network interaction occurs before the flow reaches the point where the import would be rejected by policy. Although the actual image download is blocked by the project restriction, an authenticated user can coerce the daemon into making blind HEAD requests to arbitrary destinations. These requests include server metadata in custom headers (Incus-Server-Architectures, Incus-Server-Version), which discloses information about the host environment to the attacker-controlled endpoint. This blind SSRF primitive can be used to probe internal services, unroutable address space, or cloud metadata endpoints reachable from the host. This vulnerability pattern is similar to CVE-2026-24767. This issue has been fixed in version 7.0.0.

Kata Containers is an open source project focusing on a standard implementation of lightweight Virtual Machines (VMs) that perform like containers. From v3.4.0 to v3.28.0, an oversight in the CopyFile policy (and perhaps the CopyFile handler) allows untrusted hosts to write to arbitrary locations inside the guest workload image. This can be used to overwrite binaries inside the guest and exfiltrate data from containers; even those running inside CVMs. This vulnerability is fixed in v3.29.0.

Aardvark-dns is an authoritative dns server for A/AAAA container records. From 1.16.0 to 1.17.0, a truncated TCP DNS query followed by a connection reset causes aardvark-dns to enter an unrecoverable infinite error loop at 100% CPU. This vulnerability is fixed in 1.17.1.

Incus is a system container and virtual machine manager. Incus instances have an option to provide credentials to systemd in the guest. For containers, this is handled through a shared directory. Prior to version 6.23.0, an attacker can set a configuration key named something like `systemd.credential.../../../../../../root/.bashrc` to cause Incus to write outside of the `credentials` directory associated with the container. This makes use of the fact that the Incus syntax for such credentials is `systemd.credential.XYZ` where `XYZ` can itself contain more periods. While it's not possible to read any data this way, it's possible to write to arbitrary files as root, enabling both privilege escalation and denial of service attacks. Version 6.23.0 fixes the issue.

Incus is a system container and virtual machine manager. Prior to version 6.23.0, instance template files can be used to cause arbitrary read or writes as root on the host server. Incus allows for pongo2 templates within instances which can be used at various times in the instance lifecycle to template files inside of the instance. This particular implementation of pongo2 within Incus allowed for file read/write but with the expectation that the pongo2 chroot feature would isolate all such access to the instance's filesystem. This was allowed such that a template could theoretically read a file and then generate a new version of said file. Unfortunately the chroot isolation mechanism is entirely skipped by pongo2 leading to easy access to the entire system's filesystem with root privileges. Version 6.23.0 patches the issue.

Kata Containers is an open source project focusing on a standard implementation of lightweight Virtual Machines (VMs) that perform like containers. In versions prior to 3.27.0, an issue in Kata with Cloud Hypervisor allows a user of the container to modify the file system used by the Guest micro VM ultimately achieving arbitrary code execution as root in said VM. The current understanding is this doesn’t impact the security of the Host or of other containers / VMs running on that Host (note that arm64 QEMU lacks NVDIMM read-only support: It is believed that until the upstream QEMU gains this capability, a guest write could reach the image file). Version 3.27.0 patches the issue.

Kata Containers is an open source project focusing on a standard implementation of lightweight Virtual Machines (VMs) that perform like containers. In versions prior to 3.26.0, when a container image is malformed or contains no layers, containerd falls back to bind-mounting an empty snapshotter directory for the container rootfs. When the Kata runtime attempts to mount the container rootfs, the bind mount causes the rootfs to be detected as a block device, leading to the underlying device being hotplugged to the guest. This can cause filesystem-level errors on the host due to double inode allocation, and may lead to the host's block device being mounted as read-only. Version 3.26.0 contains a patch for the issue.

Incus is a system container and virtual machine manager. Versions 6.21.0 and below allow a user with the ability to launch a container with a custom image (e.g a member of the ‘incus’ group) to use directory traversal or symbolic links in the templating functionality to achieve host arbitrary file read, and host arbitrary file write. This ultimately results in arbitrary command execution on the host. When using an image with a metadata.yaml containing templates, both the source and target paths are not checked for symbolic links or directory traversal. This can also be exploited in IncusOS. A fix is planned for versions 6.0.6 and 6.21.0, but they have not been released at the time of publication.

Incus is a system container and virtual machine manager. An issue in versions prior to 6.0.6 and 6.19.0 affects any Incus user in an environment where an unprivileged user may have root access to a container with an attached custom storage volume that has the `security.shifted` property set to `true` as well as access to the host as an unprivileged user. The most common case for this would be systems using `incus-user` with the less privileged `incus` group to provide unprivileged users with an isolated restricted access to Incus. Such users may be able to create a custom storage volume with the necessary property (depending on kernel and filesystem support) and can then write a setuid binary from within the container which can be executed as an unprivileged user on the host to gain root privileges. A patch for this issue is expected in versions 6.0.6 and 6.19.0. As a workaround, permissions can be manually restricted until a patched version of Incus is deployed.

A flaw was found in Go. When FIPS mode is enabled on a system, container runtimes may incorrectly handle certain file paths due to improper validation in the containers/common Go library. This flaw allows an attacker to exploit symbolic links and trick the system into mounting sensitive host directories inside a container. This issue also allows attackers to access critical host files, bypassing the intended isolation between containers and the host system.

A flaw was found in Aardvark-dns, which is vulnerable to a Denial of Service attack due to the serial processing of TCP DNS queries. An attacker can exploit this flaw by keeping a TCP connection open indefinitely, causing the server to become unresponsive and resulting in other DNS queries timing out. This issue prevents legitimate users from accessing DNS services, thereby disrupting normal operations and causing service downtime.

lxc-user-nic in lxc through 5.0.1 is installed setuid root, and may allow local users to infer whether any file exists, even within a protected directory tree, because "Failed to open" often indicates that a file does not exist, whereas "does not refer to a network namespace path" often indicates that a file exists. NOTE: this is different from CVE-2018-6556 because the CVE-2018-6556 fix design was based on the premise that "we will report back to the user that the open() failed but the user has no way of knowing why it failed"; however, in many realistic cases, there are no plausible reasons for failing except that the file does not exist.

A malicious container image can consume an unbounded amount of memory when being pulled to a container runtime host, such as Red Hat Enterprise Linux using podman, or OpenShift Container Platform. An attacker can use this flaw to trick a user, with privileges to pull container images, into crashing the process responsible for pulling the image. This flaw affects containers-image versions before 5.2.0.

An issue was discovered in the containers crate before 0.9.11 for Rust. When a panic occurs, a util::{mutate,mutate2} double drop can be performed.

An issue was discovered in Kata Containers through 1.11.3 and 2.x through 2.0-rc1. The runtime will execute binaries given using annotations without any kind of validation. Someone who is granted access rights to a cluster will be able to have kata-runtime execute arbitrary binaries as root on the worker nodes.

An improper file permissions vulnerability affects Kata Containers prior to 1.11.5. When using a Kubernetes hostPath volume and mounting either a file or directory into a container as readonly, the file/directory is mounted as readOnly inside the container, but is still writable inside the guest. For a container breakout situation, a malicious guest can potentially modify or delete files/directories expected to be read-only.

A malicious guest compromised before a container creation (e.g. a malicious guest image or a guest running multiple containers) can trick the kata runtime into mounting the untrusted container filesystem on any host path, potentially allowing for code execution on the host. This issue affects: Kata Containers 1.11 versions earlier than 1.11.1; Kata Containers 1.10 versions earlier than 1.10.5; Kata Containers 1.9 and earlier versions.

Kata Containers before 1.11.0 on Cloud Hypervisor persists guest filesystem changes to the underlying image file on the host. A malicious guest can overwrite the image file to gain control of all subsequent guest VMs. Since Kata Containers uses the same VM image file with all VMMs, this issue may also affect QEMU and Firecracker based guests.

In LXC 2.0, many template scripts download code over cleartext HTTP, and omit a digital-signature check, before running it to bootstrap containers.

runc through 1.0-rc6, as used in Docker before 18.09.2 and other products, allows attackers to overwrite the host runc binary (and consequently obtain host root access) by leveraging the ability to execute a command as root within one of these types of containers: (1) a new container with an attacker-controlled image, or (2) an existing container, to which the attacker previously had write access, that can be attached with docker exec. This occurs because of file-descriptor mishandling, related to /proc/self/exe.

lxc-user-nic when asked to delete a network interface will unconditionally open a user provided path. This code path may be used by an unprivileged user to check for the existence of a path which they wouldn't otherwise be able to reach. It may also be used to trigger side effects by causing a (read-only) open of special kernel files (ptmx, proc, sys). Affected releases are LXC: 2.0 versions above and including 2.0.9; 3.0 versions above and including 3.0.0, prior to 3.0.2.

lxc-attach in LXC before 1.0.9 and 2.x before 2.0.6 allows an attacker inside of an unprivileged container to use an inherited file descriptor, of the host's /proc, to access the rest of the host's filesystem via the openat() family of syscalls.

lxc-user-nic in Linux Containers (LXC) allows local users with a lxc-usernet allocation to create network interfaces on the host and choose the name of those interfaces by leveraging lack of netns ownership check.

An issue was discovered in Linux Containers (LXC) before 2016-02-22. When executing a program via lxc-attach, the nonpriv session can escape to the parent session by using the TIOCSTI ioctl to push characters into the terminal's input buffer, allowing an attacker to escape the container.