Kubernetes before 1.2.0-alpha.5 allows remote attackers to read arbitrary pod logs via a container name.

Kubernetes in OpenShift3 allows remote authenticated users to use the private images of other users should they know the name of said image.

The API server in Kubernetes does not properly check admission control, which allows remote authenticated users to access additional resources via a crafted patched object.

Openshift allows remote attackers to gain privileges by updating a build configuration that was created with an allowed type to a type that is not allowed.

It was found that Kubernetes as used by Openshift Enterprise 3 did not correctly validate X.509 client intermediate certificate host name fields. An attacker could use this flaw to bypass authentication requirements by using a specially crafted X.509 certificate.

Kubernetes version 1.5.0-1.5.4 is vulnerable to a privilege escalation in the PodSecurityPolicy admission plugin resulting in the ability to make use of any existing PodSecurityPolicy object.

Default access permissions for Persistent Volumes (PVs) created by the Kubernetes Azure cloud provider in versions 1.6.0 to 1.6.5 are set to "container" which exposes a URI that can be accessed without authentication on the public internet. Access to the URI string requires privileged access to the Kubernetes cluster or authenticated access to the Azure portal.

In Kubernetes versions 1.3.x, 1.4.x, 1.5.x, 1.6.x and prior to versions 1.7.14, 1.8.9 and 1.9.4 containers using subpath volume mounts with any volume type (including non-privileged pods, subject to file permissions) can access files/directories outside of the volume, including the host's filesystem.

In Kubernetes versions 1.3.x, 1.4.x, 1.5.x, 1.6.x and prior to versions 1.7.14, 1.8.9 and 1.9.4 containers using a secret, configMap, projected or downwardAPI volume can trigger deletion of arbitrary files/directories from the nodes where they are running.

Kubernetes CRI-O version prior to 1.9 contains a Privilege Context Switching Error (CWE-270) vulnerability in the handling of ambient capabilities that can result in containers running with elevated privileges, allowing users abilities they should not have. This attack appears to be exploitable via container execution. This vulnerability appears to have been fixed in 1.9.

In Kubernetes versions 1.5.x, 1.6.x, 1.7.x, 1.8.x, and prior to version 1.9.6, the kubectl cp command insecurely handles tar data returned from the container, and can be caused to overwrite arbitrary local files.

In Kubernetes versions 1.9.0-1.9.9, 1.10.0-1.10.5, and 1.11.0-1.11.1, user input was handled insecurely while setting up volume mounts on Windows nodes, which could lead to command line argument injection.

Improper validation of URL redirection in the Kubernetes API server in versions prior to v1.14.0 allows an attacker-controlled Kubelet to redirect API server requests from streaming endpoints to arbitrary hosts. Impacted API servers will follow the redirect as a GET request with client-certificate credentials for authenticating to the Kubelet.

In Minikube versions 0.3.0-0.29.0, minikube exposes the Kubernetes Dashboard listening on the VM IP at port 30000. In VM environments where the IP is easy to predict, the attacker can use DNS rebinding to indirectly make requests to the Kubernetes Dashboard, create a new Kubernetes Deployment running arbitrary code. If minikube mount is in use, the attacker could also directly access the host filesystem.

Versions < 1.5 of the Kubernetes ingress default backend, which handles invalid ingress traffic, exposed prometheus metrics publicly.

In all Kubernetes versions prior to v1.10.11, v1.11.5, and v1.12.3, incorrect handling of error responses to proxied upgrade requests in the kube-apiserver allowed specially crafted requests to establish a connection through the Kubernetes API server to backend servers, then send arbitrary requests over the same connection directly to the backend, authenticated with the Kubernetes API server's TLS credentials used to establish the backend connection.

Kubernetes Dashboard before 1.10.1 allows attackers to bypass authentication and use Dashboard's Service Account for reading secrets within the cluster.

In all Kubernetes versions prior to v1.11.8, v1.12.6, and v1.13.4, users that are authorized to make patch requests to the Kubernetes API Server can send a specially crafted patch of type "json-patch" (e.g. `kubectl patch --type json` or `"Content-Type: application/json-patch+json"`) that consumes excessive resources while processing, causing a Denial of Service on the API Server.

The kubectl cp command allows copying files between containers and the user machine. To copy files from a container, Kubernetes creates a tar inside the container, copies it over the network, and kubectl unpacks it on the user’s machine. If the tar binary in the container is malicious, it could run any code and output unexpected, malicious results. An attacker could use this to write files to any path on the user’s machine when kubectl cp is called, limited only by the system permissions of the local user. The untar function can both create and follow symbolic links. The issue is resolved in kubectl v1.11.9, v1.12.7, v1.13.5, and v1.14.0.

A security issue was discovered in the kube-state-metrics versions v1.7.0 and v1.7.1. An experimental feature was added to the v1.7.0 release that enabled annotations to be exposed as metrics. By default, the kube-state-metrics metrics only expose metadata about Secrets. However, a combination of the default `kubectl` behavior and this new feature can cause the entire secret content to end up in metric labels thus inadvertently exposing the secret content in metrics. This feature has been reverted and released as the v1.7.2 release. If you are running the v1.7.0 or v1.7.1 release, please upgrade to the v1.7.2 release as soon as possible.

In Kubernetes v1.12.0-v1.12.4 and v1.13.0, the rest.AnonymousClientConfig() method returns a copy of the provided config, with credentials removed (bearer token, username/password, and client certificate/key data). In the affected versions, rest.AnonymousClientConfig() did not effectively clear service account credentials loaded using rest.InClusterConfig()

In Kubernetes v1.8.x-v1.14.x, schema info is cached by kubectl in the location specified by --cache-dir (defaulting to $HOME/.kube/http-cache), written with world-writeable permissions (rw-rw-rw-). If --cache-dir is specified and pointed at a different location accessible to other users/groups, the written files may be modified by other users/groups and disrupt the kubectl invocation.

In kubelet v1.13.6 and v1.14.2, containers for pods that do not specify an explicit runAsUser attempt to run as uid 0 (root) on container restart, or if the image was previously pulled to the node. If the pod specified mustRunAsNonRoot: true, the kubelet will refuse to start the container as root. If the pod did not specify mustRunAsNonRoot: true, the kubelet will run the container as uid 0.

The kubectl cp command allows copying files between containers and the user machine. To copy files from a container, Kubernetes runs tar inside the container to create a tar archive, copies it over the network, and kubectl unpacks it on the user’s machine. If the tar binary in the container is malicious, it could run any code and output unexpected, malicious results. An attacker could use this to write files to any path on the user’s machine when kubectl cp is called, limited only by the system permissions of the local user. Kubernetes affected versions include versions prior to 1.12.9, versions prior to 1.13.6, versions prior to 1.14.2, and versions 1.1, 1.2, 1.4, 1.4, 1.5, 1.6, 1.7, 1.8, 1.9, 1.10, 1.11.

The Kubernetes kube-apiserver mistakenly allows access to a cluster-scoped custom resource if the request is made as if the resource were namespaced. Authorizations for the resource accessed in this manner are enforced using roles and role bindings within the namespace, meaning that a user with access only to a resource in one namespace could create, view update or delete the cluster-scoped resource (according to their namespace role privileges). Kubernetes affected versions include versions prior to 1.13.9, versions prior to 1.14.5, versions prior to 1.15.2, and versions 1.7, 1.8, 1.9, 1.10, 1.11, 1.12.

The debugging endpoint /debug/pprof is exposed over the unauthenticated Kubelet healthz port. The go pprof endpoint is exposed over the Kubelet's healthz port. This debugging endpoint can potentially leak sensitive information such as internal Kubelet memory addresses and configuration, or for limited denial of service. Versions prior to 1.15.0, 1.14.4, 1.13.8, and 1.12.10 are affected. The issue is of medium severity, but not exposed by the default configuration.

The kubectl cp command allows copying files between containers and the user machine. To copy files from a container, Kubernetes runs tar inside the container to create a tar archive, copies it over the network, and kubectl unpacks it on the user’s machine. If the tar binary in the container is malicious, it could run any code and output unexpected, malicious results. An attacker could use this to write files to any path on the user’s machine when kubectl cp is called, limited only by the system permissions of the local user. Kubernetes affected versions include versions prior to 1.13.9, versions prior to 1.14.5, versions prior to 1.15.2, and versions 1.1, 1.2, 1.4, 1.4, 1.5, 1.6, 1.7, 1.8, 1.9, 1.10, 1.11, 1.12.

The Kubernetes client-go library logs request headers at verbosity levels of 7 or higher. This can disclose credentials to unauthorized users via logs or command output. Kubernetes components (such as kube-apiserver) prior to v1.16.0, which make use of basic or bearer token authentication, and run at high verbosity levels, are affected.

The Kubernetes kubectl cp command in versions 1.1-1.12, and versions prior to 1.13.11, 1.14.7, and 1.15.4 allows a combination of two symlinks provided by tar output of a malicious container to place a file outside of the destination directory specified in the kubectl cp invocation. This could be used to allow an attacker to place a nefarious file using a symlink, outside of the destination tree.

The Kubernetes kube-controller-manager in versions v1.0-v1.17 is vulnerable to a credential leakage via error messages in mount failure logs and events for AzureFile and CephFS volumes.

Improper input validation in the Kubernetes API server in versions v1.0-1.12 and versions prior to v1.13.12, v1.14.8, v1.15.5, and v1.16.2 allows authorized users to send malicious YAML or JSON payloads, causing the API server to consume excessive CPU or memory, potentially crashing and becoming unavailable. Prior to v1.14.0, default RBAC policy authorized anonymous users to submit requests that could trigger this vulnerability. Clusters upgraded from a version prior to v1.14.0 keep the more permissive policy by default for backwards compatibility.

The Kubernetes API Server component in versions 1.1-1.14, and versions prior to 1.15.10, 1.16.7 and 1.17.3 allows an authorized user who sends malicious YAML payloads to cause the kube-apiserver to consume excessive CPU cycles while parsing YAML.

Improper input validation in Kubernetes CSI sidecar containers for external-provisioner (<v0.4.3, <v1.0.2, v1.1, <v1.2.2, <v1.3.1), external-snapshotter (<v0.4.2, <v1.0.2, v1.1, <1.2.2), and external-resizer (v0.1, v0.2) could result in unauthorized PersistentVolume data access or volume mutation during snapshot, restore from snapshot, cloning and resizing operations.

A flaw was found in cri-o, as a result of all pod-related processes being placed in the same memory cgroup. This can result in container management (conmon) processes being killed if a workload process triggers an out-of-memory (OOM) condition for the cgroup. An attacker could abuse this flaw to get host network access on an cri-o host.

Cloud Native Computing Foundation (CNCF) CNI (Container Networking Interface) 0.7.4 has a network firewall misconfiguration which affects Kubernetes. The CNI 'portmap' plugin, used to setup HostPorts for CNI, inserts rules at the front of the iptables nat chains; which take precedence over the KUBE- SERVICES chain. Because of this, the HostPort/portmap rule could match incoming traffic even if there were better fitting, more specific service definition rules like NodePorts later in the chain. The issue is fixed in CNI 0.7.5 and Kubernetes 1.11.9, 1.12.7, 1.13.5, and 1.14.0.

The Kubelet component in versions 1.15.0-1.15.9, 1.16.0-1.16.6, and 1.17.0-1.17.2 has been found to be vulnerable to a denial of service attack via the kubelet API, including the unauthenticated HTTP read-only API typically served on port 10255, and the authenticated HTTPS API typically served on port 10250.

The Kubernetes API server component in versions prior to 1.15.9, 1.16.0-1.16.6, and 1.17.0-1.17.2 has been found to be vulnerable to a denial of service attack via successful API requests.

The Kubernetes ingress-nginx component prior to version 0.28.0 allows a user with the ability to create namespaces and to read and create ingress objects to overwrite the password file of another ingress which uses nginx.ingress.kubernetes.io/auth-type: basic and which has a hyphenated namespace or secret name.

Kubernetes API server in all versions allow an attacker who is able to create a ClusterIP service and set the spec.externalIPs field, to intercept traffic to that IP address. Additionally, an attacker who is able to patch the status (which is considered a privileged operation and should not typically be granted to users) of a LoadBalancer service can set the status.loadBalancer.ingress.ip to similar effect.

The Kubernetes kube-controller-manager in versions v1.0-1.14, versions prior to v1.15.12, v1.16.9, v1.17.5, and version v1.18.0 are vulnerable to a Server Side Request Forgery (SSRF) that allows certain authorized users to leak up to 500 bytes of arbitrary information from unprotected endpoints within the master's host network (such as link-local or loopback services).

The Kubernetes kubelet component in versions 1.1-1.16.12, 1.17.0-1.17.8 and 1.18.0-1.18.5 do not account for disk usage by a pod which writes to its own /etc/hosts file. The /etc/hosts file mounted in a pod by kubelet is not included by the kubelet eviction manager when calculating ephemeral storage usage by a pod. If a pod writes a large amount of data to the /etc/hosts file, it could fill the storage space of the node and cause the node to fail.

The Kubelet and kube-proxy components in versions 1.1.0-1.16.10, 1.17.0-1.17.6, and 1.18.0-1.18.3 were found to contain a security issue which allows adjacent hosts to reach TCP and UDP services bound to 127.0.0.1 running on the node or in the node's network namespace. Such a service is generally thought to be reachable only by other processes on the same host, but due to this defeect, could be reachable by other hosts on the same LAN as the node, or by containers running on the same node as the service.

The Kubernetes kube-apiserver in versions v1.6-v1.15, and versions prior to v1.16.13, v1.17.9 and v1.18.6 are vulnerable to an unvalidated redirect on proxied upgrade requests that could allow an attacker to escalate privileges from a node compromise to a full cluster compromise.

A security issue was discovered in Kubernetes where actors that control the responses of MutatingWebhookConfiguration or ValidatingWebhookConfiguration requests are able to redirect kube-apiserver requests to private networks of the apiserver. If that user can view kube-apiserver logs when the log level is set to 10, they can view the redirected responses and headers in the logs.

As mitigations to a report from 2019 and CVE-2020-8555, Kubernetes attempts to prevent proxied connections from accessing link-local or localhost networks when making user-driven connections to Services, Pods, Nodes, or StorageClass service providers. As part of this mitigation Kubernetes does a DNS name resolution check and validates that response IPs are not in the link-local (169.254.0.0/16) or localhost (127.0.0.0/8) range. Kubernetes then performs a second DNS resolution without validation for the actual connection. If a non-standard DNS server returns different non-cached responses, a user may be able to bypass the proxy IP restriction and access private networks on the control plane.

In Kubernetes clusters using VSphere as a cloud provider, with a logging level set to 4 or above, VSphere cloud credentials will be leaked in the cloud controller manager's log. This affects < v1.19.3.

In Kubernetes clusters using a logging level of at least 4, processing a malformed docker config file will result in the contents of the docker config file being leaked, which can include pull secrets or other registry credentials. This affects < v1.19.3, < v1.18.10, < v1.17.13.

In Kubernetes, if the logging level is set to at least 9, authorization and bearer tokens will be written to log files. This can occur both in API server logs and client tool output like kubectl. This affects <= v1.19.3, <= v1.18.10, <= v1.17.13, < v1.20.0-alpha2.

In Kubernetes clusters using Ceph RBD as a storage provisioner, with logging level of at least 4, Ceph RBD admin secrets can be written to logs. This occurs in kube-controller-manager's logs during provisioning of Ceph RBD persistent claims. This affects < v1.19.3, < v1.18.10, < v1.17.13.

Kubernetes Secrets Store CSI Driver versions v0.0.15 and v0.0.16 allow an attacker who can modify a SecretProviderClassPodStatus/Status resource the ability to write content to the host filesystem and sync file contents to Kubernetes Secrets. This includes paths under var/lib/kubelet/pods that contain other Kubernetes Secrets.

Kubernetes CSI snapshot-controller prior to v2.1.3 and v3.0.2 could panic when processing a VolumeSnapshot custom resource when: - The VolumeSnapshot referenced a non-existing PersistentVolumeClaim and the VolumeSnapshot did not reference any VolumeSnapshotClass. - The snapshot-controller crashes, is automatically restarted by Kubernetes, and processes the same VolumeSnapshot custom resource after the restart, entering an endless crashloop. Only the volume snapshot feature is affected by this vulnerability. When exploited, users can’t take snapshots of their volumes or delete the snapshots. All other Kubernetes functionality is not affected.

Kubernetes Java client libraries in version 10.0.0 and versions prior to 9.0.1 allow writes to paths outside of the current directory when copying multiple files from a remote pod which sends a maliciously crafted archive. This can potentially overwrite any files on the system of the process executing the client code.

A security issue was discovered in kube-apiserver that could allow node updates to bypass a Validating Admission Webhook. Clusters are only affected by this vulnerability if they run a Validating Admission Webhook for Nodes that denies admission based at least partially on the old state of the Node object. Validating Admission Webhook does not observe some previous fields.

Kube-proxy on Windows can unintentionally forward traffic to local processes listening on the same port (“spec.ports[*].port”) as a LoadBalancer Service when the LoadBalancer controller does not set the “status.loadBalancer.ingress[].ip” field. Clusters where the LoadBalancer controller sets the “status.loadBalancer.ingress[].ip” field are unaffected.

A security issue was discovered in Kubernetes where a user may be able to redirect pod traffic to private networks on a Node. Kubernetes already prevents creation of Endpoint IPs in the localhost or link-local range, but the same validation was not performed on EndpointSlice IPs.

Loading specially-crafted yaml with the Kubernetes Java Client library can lead to code execution.

A security issue was discovered with Kubernetes that could enable users to send network traffic to locations they would otherwise not have access to via a confused deputy attack.

A security issue was discovered in Kubernetes where a user may be able to create a container with subpath volume mounts to access files & directories outside of the volume, including on the host filesystem.

A security issue was discovered in ingress-nginx where a user that can create or update ingress objects can use the custom snippets feature to obtain all secrets in the cluster.

kubectl does not neutralize escape, meta or control sequences contained in the raw data it outputs to a terminal. This includes but is not limited to the unstructured string fields in objects such as Events.

A security issue was discovered in ingress-nginx where a user that can create or update ingress objects can use the spec.rules[].http.paths[].path field of an Ingress object (in the networking.k8s.io or extensions API group) to obtain the credentials of the ingress-nginx controller. In the default configuration, that credential has access to all secrets in the cluster.

A security issue was discovered in ingress-nginx where a user that can create or update ingress objects can use .metadata.annotations in an Ingress object (in the networking.k8s.io or extensions API group) to obtain the credentials of the ingress-nginx controller. In the default configuration, that credential has access to all secrets in the cluster.

A security issue was discovered in ingress-nginx where a user that can create or update ingress objects can use a newline character to bypass the sanitization of the `spec.rules[].http.paths[].path` field of an Ingress object (in the `networking.k8s.io` or `extensions` API group) to obtain the credentials of the ingress-nginx controller. In the default configuration, that credential has access to all secrets in the cluster.

Windows workloads can run as ContainerAdministrator even when those workloads set the runAsNonRoot option to true.

An incorrect sysctls validation vulnerability was found in CRI-O 1.18 and earlier. The sysctls from the list of "safe" sysctls specified for the cluster will be applied to the host if an attacker is able to create a pod with a hostIPC and hostNetwork kernel namespace.

A flaw was found in CRI-O in the way it set kernel options for a pod. This issue allows anyone with rights to deploy a pod on a Kubernetes cluster that uses the CRI-O runtime to achieve a container escape and arbitrary code execution as root on the cluster node, where the malicious pod was deployed.

A vulnerability was found in CRI-O that causes memory or disk space exhaustion on the node for anyone with access to the Kube API. The ExecSync request runs commands in a container and logs the output of the command. This output is then read by CRI-O after command execution, and it is read in a manner where the entire file corresponding to the output of the command is read in. Thus, if the output of the command is large it is possible to exhaust the memory or the disk space of the node when CRI-O reads the output of the command. The highest threat from this vulnerability is system availability.

A security issue was discovered in aws-iam-authenticator where an allow-listed IAM identity may be able to modify their username and escalate privileges.

A flaw was found in cri-o, where containers were incorrectly started with non-empty default permissions. A vulnerability was found in Moby (Docker Engine) where containers started incorrectly with non-empty inheritable Linux process capabilities. This flaw allows an attacker with access to programs with inheritable file capabilities to elevate those capabilities to the permitted set when execve(2) runs.

Incorrect handling of the supplementary groups in the CRI-O container engine might lead to sensitive information disclosure or possible data modification if an attacker has direct access to the affected container where supplementary groups are used to set access permissions and is able to execute a binary code in that container.

Users authorized to list or watch one type of namespaced custom resource cluster-wide can read custom resources of a different type in the same API group without authorization. Clusters are impacted by this vulnerability if all of the following are true: 1. There are 2+ CustomResourceDefinitions sharing the same API group 2. Users have cluster-wide list or watch authorization on one of those custom resources. 3. The same users are not authorized to read another custom resource in the same API group.

A security issue was discovered in kube-apiserver that allows an aggregated API server to redirect client traffic to any URL. This could lead to the client performing unexpected actions as well as forwarding the client's API server credentials to third parties.

Users may have access to secure endpoints in the control plane network. Kubernetes clusters are only affected if an untrusted user can modify Node objects and send proxy requests to them. Kubernetes supports node proxying, which allows clients of kube-apiserver to access endpoints of a Kubelet to establish connections to Pods, retrieve container logs, and more. While Kubernetes already validates the proxying address for Nodes, a bug in kube-apiserver made it possible to bypass this validation. Bypassing this validation could allow authenticated requests destined for Nodes to to the API server's private network.

The version of cri-o as released for Red Hat OpenShift Container Platform 4.9.48, 4.10.31, and 4.11.6 via RHBA-2022:6316, RHBA-2022:6257, and RHBA-2022:6658, respectively, included an incorrect version of cri-o missing the fix for CVE-2022-27652, which was previously fixed in OCP 4.9.41 and 4.10.12 via RHBA-2022:5433 and RHSA-2022:1600. This issue could allow an attacker with access to programs with inheritable file capabilities to elevate those capabilities to the permitted set when execve(2) runs. For more details, see https://access.redhat.com/security/cve/CVE-2022-27652.

A vulnerability was found in cri-o. This issue allows the addition of arbitrary lines into /etc/passwd by use of a specially crafted environment variable.

Ingress-nginx `path` sanitization can be bypassed with `log_format` directive.

This vulnerability exposes a network port in minikube running on macOS with Docker driver that could enable unexpected remote access to the minikube container.

An authentication bypass vulnerability was discovered in kube-apiserver. This issue could allow a remote, authenticated attacker who has been given permissions "update, patch" the "pods/ephemeralcontainers" subresource beyond what the default is. They would then need to create a new pod or patch one that they already have access to. This might allow evasion of SCC admission restrictions, thereby gaining control of a privileged pod.

Privilege Escalation in kOps using GCE/GCP Provider in Gossip Mode.

This vulnerability enables ssh access to minikube container using a default password.

A security issue was discovered in Kubelet that allows pods to bypass the seccomp profile enforcement. Pods that use localhost type for seccomp profile but specify an empty profile field, are affected by this issue. In this scenario, this vulnerability allows the pod to run in unconfined (seccomp disabled) mode. This bug affects Kubelet.

Users may be able to launch containers using images that are restricted by ImagePolicyWebhook when using ephemeral containers. Kubernetes clusters are only affected if the ImagePolicyWebhook admission plugin is used together with ephemeral containers.

Users may be able to launch containers that bypass the mountable secrets policy enforced by the ServiceAccount admission plugin when using ephemeral containers. The policy ensures pods running with a service account may only reference secrets specified in the service account’s secrets field. Kubernetes clusters are only affected if the ServiceAccount admission plugin and the `kubernetes.io/enforce-mountable-secrets` annotation are used together with ephemeral containers.

Kubernetes secrets-store-csi-driver in versions before 1.3.3 discloses service account tokens in logs.

A security issue was discovered in Kubernetes where a user that can create pods on Windows nodes may be able to escalate to admin privileges on those nodes. Kubernetes clusters are only affected if they include Windows nodes.

A security issue was discovered in Kubernetes where a user that can create pods on Windows nodes running kubernetes-csi-proxy may be able to escalate to admin privileges on those nodes. Kubernetes clusters are only affected if they include Windows nodes running kubernetes-csi-proxy.

A security issue was discovered in Kubernetes where a user that can create pods on Windows nodes may be able to escalate to admin privileges on those nodes. Kubernetes clusters are only affected if they include Windows nodes.

Ingress nginx annotation injection causes arbitrary command execution.

Code injection via nginx.ingress.kubernetes.io/permanent-redirect annotation.

A security issue was discovered in Kubernetes where a user that can create pods and persistent volumes on Windows nodes may be able to escalate to admin privileges on those nodes. Kubernetes clusters are only affected if they are using an in-tree storage plugin for Windows nodes.

A flaw was found in cri-o. A malicious container can create a symbolic link to arbitrary files on the host via directory traversal (“../“). This flaw allows the container to read and write to arbitrary files on the host system.