threat-model.md

Kata Containers threat model

This document discusses threat models associated with the Kata Containers project. Kata was designed to provide additional isolation of container workloads, protecting the host infrastructure from potentially malicious container users or workloads. Since Kata Containers adds a level of isolation on top of traditional containers, the focus is on the additional layer provided, not on traditional container security.

This document provides a brief background on containers and layered security, describes the interface to Kata from CRI runtimes, a review of utilized virtual machine interfaces, and then a review of threats.

Kata security objective

Kata seeks to prevent an untrusted container workload or user of that container workload to gain control of, obtain information from, or tamper with the host infrastructure.

In our scenario, an asset is anything on the host system, or elsewhere in the cluster infrastructure. The attacker is assumed to be either a malicious user or the workload itself running within the container. The goal of Kata is to prevent attacks which would allow any access to the defined assets.

Background on containers, layered security

Traditional containers leverage several key Linux kernel features to provide isolation and a view that the container workload is the only entity running on the host. Key features include Namespaces, cgroups, capablities, SELinux and seccomp. The canonical runtime for creating such a container is runc. In the remainder of the document, the term traditional-container will be used to describe a container workload created by runc.

Kata Containers provides a second layer of isolation on top of those provided by traditional-containers. The hardware virtualization interface is the basis of this additional layer. Kata launches a lightweight virtual machine, and uses the guest’s Linux kernel to create a container workload, or workloads in the case of multi-container pods. In Kubernetes and in the Kata implementation, the sandbox is carried out at the pod level. In Kata, this sandbox is created using a virtual machine.

Interface to Kata Containers: CRI, v2-shim, OCI

A typical Kata Containers deployment uses Kubernetes with a CRI implementation. On every node, Kubelet will interact with a CRI implementor, which will in turn interface with an OCI based runtime, such as Kata Containers. Typical CRI implementors are cri-o and containerd.

The CRI API, as defined at the Kubernetes CRI-API repo, results in a few constructs being supported by the CRI implementation, and ultimately in the OCI runtime creating the workloads.

In order to run a container inside of the Kata sandbox, several virtual machine devices and interfaces are required. Kata translates sandbox and container definitions to underlying virtualization technologies provided by a set of virtual machine monitors (VMMs) and hypervisors. These devices and their underlying implementations are discussed in detail in the following section.

Interface to the Kata sandbox/virtual machine

In case of Kata, today the devices which we need in the guest are:

• Storage: In the current design of Kata Containers, we are reliant on the CRI implementor to assist in image handling and volume management on the host. As a result, we need to support a way of passing to the sandbox the container rootfs, volumes requested by the workload, and any other volumes created to facilitate sharing of secrets and configmaps with the containers. Depending on how these are managed, a block based device or file-system sharing is required. Kata Containers does this by way of virtio-blk and/or virtio-fs.
• Networking: A method for enabling network connectivity with the workload is required. Typically this will be done providing a TAP device to the VMM, and this will be exposed to the guest as a virtio-net device. It is feasible to pass in a NIC device directly, in which case VFIO is leveraged and the device itself will be exposed to the guest.
• Control: In order to interact with the guest agent and retrieve STDIO from containers, a medium of communication is required. This is available via virtio-vsock.
• Devices: VFIO is utilized when devices are passed directly to the virtual machine and exposed to the container.
• Dynamic Resource Management: ACPI is utilized to allow for dynamic VM resource management (for example: CPU, memory, device hotplug). This is required when containers are resized, or more generally when containers are added to a pod.

How these devices are utilized varies depending on the VMM utilized. We clarify the default settings provided when integrating Kata with the QEMU, Dragonball, Firecracker and Cloud Hypervisor VMMs in the following sections.

Virtual Machine Monitor(s)

In a KVM/QEMU (any other VMM utilizing KVM) virtualization setup, all virtual machines (VMs) share the same host kernel. This shared environment can lead to scenarios where one VM could potentially impact the performance or stability of other VMs, including the possibility of a Denial of Service attack.

• Kernel Vulnerabilities: Since all VMs rely on the host's kernel, a vulnerability in the kernel could be exploited by a process running within one VM to affect the entire system. This could lead to scenarios where the compromised VM impacts other VMs or even takes down the host.

• Improper Isolation and Containment: If the virtualization environment is not correctly configured, processes in one VM might impact other VMs. This could occur through improper isolation of network traffic, shared file systems, or other inter-VM communication channels.

• Hypervisor Vulnerabilities: Flaws in the KVM hypervisor or QEMU could be exploited to cause information disclosure, data tampering, elevation of privileges, denial of service, and others. Since KVM/QEMU leverages the host kernel for its operation, any exploit at this level can have widespread impacts.

• Malicious or Flawed Guest Operating Systems: A guest operating system that is maliciously designed or has serious flaws could engage in activities that disrupt the normal operation of the host or other guests. This might include aggressive network activity or interactions with the virtualization stack that lead to instability.

• Resource Exhaustion: A VM could consume excessive shared resources such as CPU, memory, or I/O bandwidth, leading to resource starvation for other VMs. This could be due to misconfiguration, a runaway process, or a deliberate denial of service attack from a compromised VM.

Devices

Each virtio device is implemented by a backend, which may execute within userspace on the host (vhost-user), the VMM itself, or within the host kernel (vhost). While it may provide enhanced performance, vhost devices are often seen as higher risk since an exploit would be already running within the kernel space. While VMM and vhost-user are both in userspace on the host, vhost-user generally allows for the back-end process to require less system calls and capabilities compared to a full VMM.

virtio-blk and virtio-scsi

The backend for virtio-blk and virtio-scsi are based in the VMM itself (ring3 in the context of x86) by default for Cloud Hypervisor, Firecracker and QEMU. While vhost based back-ends are available for QEMU, it is not recommended. vhost-user back-ends are being added for Cloud Hypervisor, they are not utilized in Kata today.

virtio-fs

virtio-fs is supported in Cloud Hypervisor and QEMU. virtio-fs's interaction with the host filesystem is done through a vhost-user daemon, virtiofsd. The virtio-fs client, running in the guest, will generate requests to access files. virtiofsd will receive requests, open the file, and request the VMM to mmap it into the guest. When DAX is utilized, the guest will access the host's page cache, avoiding the need for copy and duplication. DAX is still an experimental feature, and is not enabled by default.

From the virtiofsd documentation: This program must be run as the root user. Upon startup the program will switch into a new file system namespace with the shared directory tree as its root. This prevents “file system escapes” due to symlinks and other file system objects that might lead to files outside the shared directory. The program also sandboxes itself using seccomp(2) to prevent ptrace(2) and other vectors that could allow an attacker to compromise the system after gaining control of the virtiofsd process.

DAX-less support for virtio-fs is available as of the 5.4 Linux kernel. QEMU VMM supports virtio-fs as of v4.2. Cloud Hypervisor supports virtio-fs.

virtio-net

virtio-net has many options, depending on the VMM and Kata configurations.

QEMU networking

While QEMU has options for vhost, virtio-net and vhost-user, the virtio-net backend for Kata defaults to vhost-net for performance reasons. The default configuration is being reevaluated.

Firecracker networking

For Firecracker, the virtio-net backend is within Firecracker's VMM.

Cloud Hypervisor networking

For Cloud Hypervisor, the current backend default is within the VMM. vhost-user-net support is being added (written in rust, Cloud Hypervisor specific).

Dragonball networking

For Dragonball, the virtio-net backend default is within Dragonbasll's VMM.

virtio-vsock

QEMU vsock

In QEMU, vsock is backed by vhost_vsock, which runs within the kernel itself.

Dragonball, Firecracker and Cloud Hypervisor

In Dragonball, Firecracker and Cloud Hypervisor, vsock is backed by a unix-domain-socket in the hosts userspace.

VFIO

Utilizing VFIO, devices can be passed through to the virtual machine. Exposure to the host is limited to gaps in device pass-through handling. This is supported in QEMU and Cloud Hypervisor, but not Firecracker.

• Device Isolation Failure: One of the primary risks associated with VFIO is the failure to isolate the physical device. If a VM can affect the operation of the physical device in a way that impacts other VMs or the host system, it could lead to security breaches or system instability.

• DMA Attacks: Direct Memory Access (DMA) attacks are a significant concern with VFIO. Since the device has direct access to the system's memory, there's a risk that a compromised VM could use its assigned device to read or write memory outside of its allocated space, potentially accessing sensitive information or affecting the host or other VMs.

• Firmware Vulnerabilities: Devices attached via VFIO rely on their firmware, which can have vulnerabilities. A compromised device firmware could be exploited to gain unauthorized access or to disrupt the system. Resource Starvation: Improperly managed, a VM with direct access to hardware resources could monopolize those resources, leading to performance degradation or denial of service for other VMs or the host system.

• Escalation of Privileges: If a VM with VFIO access is compromised, it could potentially be used to gain higher privileges than intended, especially if the I/O devices have capabilities that are not adequately controlled or monitored.

• Improper Configuration and Management: Human errors in configuring VFIO, such as incorrect group or user permissions, can expose the system to risks. Additionally, inadequate monitoring and management of the VMs and their devices can lead to security lapses.

• Software Vulnerabilities: Like any software, the components of VFIO (like the kernel modules, device drivers, and management tools) can have vulnerabilities that might be exploited by an attacker to compromise the security of the system. Inter-VM Interference and Side-Channel Attacks: Even with device assignment, there could be side-channel attacks where an attacker VM infers sensitive information from the physical device's behavior or through shared resources like cache.

ACPI (Dragonball uses Upcall)

ACPI is necessary for hotplugging of CPU, memory and devices. ACPI is available in QEMU and Cloud Hypervisor. Device, CPU and memory hotplug are not available in Firecracker.

• Hypervisor Vulnerabilities: In virtualized environments, the hypervisor manages ACPI calls for virtual machines (VMs). If the hypervisor has vulnerabilities in handling ACPI requests, it could lead to escalated privileges or other security breaches.

• VM Escape: A sophisticated attack could exploit ACPI functionality to achieve a VM escape, where malicious code in a VM breaks out to the host system or other VMs. Firmware Attacks in a Virtualized Context: Similar to physical environments, firmware-based attacks (including those targeting ACPI) in virtualized systems can be persistent and difficult to detect. In a virtualized environment, such attacks might not only compromise the host system but also all the VMs running on it.

• Resource Starvation Attacks: ACPI functionality could be exploited to manipulate power management features, causing denial of service through resource starvation. For example, an attacker could force a VM into a low-power state, degrading its performance or availability.

• Compromised VMs Affecting Host ACPI Settings: If a VM is compromised, it might be used to alter ACPI settings on the host, affecting all VMs on that host. This could lead to various impacts, from performance degradation to system instability.

• Supply Chain Risks: As with non-virtualized environments, the firmware, including ACPI firmware used in virtualized environments, could be compromised during the supply chain process, leading to vulnerabilities that affect all VMs running on the hardware.

Devices and threat model