Microsoft has discovered several vulnerabilities, collectively referred to as Nimbuspwn, that could allow an attacker to elevate privileges to root on many Linux desktop endpoints. The vulnerabilities can be chained together to gain root privileges on Linux systems, allowing attackers to deploy payloads, like a root backdoor, and perform other malicious actions via arbitrary root code execution. Moreover, the Nimbuspwn vulnerabilities could potentially be leveraged as a vector for root access by more sophisticated threats, such as malware or ransomware, to achieve greater impact on vulnerable devices.
We discovered the vulnerabilities by listening to messages on the System Bus while performing code reviews and dynamic analysis on services that run as root, noticing an odd pattern in a systemd unit called networkd-dispatcher. Reviewing the code flow for networkd-dispatcher revealed multiple security concerns, including directory traversal, symlink race, and time-of-check-time-of-use race condition issues, which could be leveraged to elevate privileges and deploy malware or carry out other malicious activities. We shared these vulnerabilities with the relevant maintainers through Coordinated Vulnerability Disclosure (CVD) via Microsoft Security Vulnerability Research (MSVR). Fixes for these vulnerabilities, now identified as CVE-2022-29799 and CVE-2022-29800, have been successfully deployed by the maintainer of the networkd-dispatcher, Clayton Craft. We wish to thank Clayton for his professionalism and collaboration in resolving those issues. Users of networkd-dispatcher are encouraged to update their instances.
As organizational environments continue to rely on a diverse range of devices and systems, they require comprehensive solutions that provide cross-platform protection and a holistic view of their security posture to mitigate threats, such as Nimbuspwn. The growing number of vulnerabilities on Linux environments emphasize the need for strong monitoring of the platform’s operating system and its components. Microsoft Defender for Endpoint enables organizations to gain this necessary visibility and detect such threats on Linux devices, allowing organizations to detect, manage, respond, and remediate vulnerabilities and threats across different platforms, including Windows, Linux, Mac, iOS, and Android.
In this blog post, we will share some information about the affected components and examine the vulnerabilities we uncovered. Detailing how our cross-domain visibility helps us uncover new and unknown threats to continually improve security, we are also sharing details from our research with the larger security community to underscore the importance of securing platforms and devices.
D-Bus (short for “Desktop-Bus”) is an inter-process communication channel (IPC) mechanism developed by the freedesktop.org project. D-Bus is a software-bus and allows processes on the same endpoint to communicate by transmitting messages and responding to them. D-Bus supports two main ways of communicating:
An example of D-Bus usage would be receiving a video chat by a popular video conferencing app–once a video is established, the video conferencing app could send a D-bus signal publishing that a call has started. Apps listening to that message could respond appropriately–for example, mute their audio.
There are many D-Bus components shipped by default on popular Linux desktop environments. Since those components run at different privileges and respond to messages, D-Bus components are an attractive target for attackers. Indeed, there have been interesting vulnerabilities in the past related to buggy D-Bus services, including USBCreator Elevation of Privilege, Blueman Elevation of Privilege by command injection, and other similar scenarios.
D-Bus exposes a global System Bus and a per-session Session Bus. From an attacker’s perspective, the System Bus is more attractive since it will commonly have services that run as root listening to it.
When connecting to the D-Bus, components are assigned with a unique identifier, which mitigates against attacks abusing PID-recycling. The unique identifier starts with a colon and has numbers in it separated by dots, such as “:1.337”. Components can use the D-Bus API to own identifiable names such as “org.freedesktop.Avahi” or “com.ubuntu.SystemService”. For D-Bus to allow such ownership, the requesting process context must be allowed under the D-Bus configuration files. Those configuration files are well documented and maintained under /usr/local/share/dbus-1/system.conf and /usr/local/share/dbus-1/session.conf (on some systems under /usr/local/dbus-1 directly). Specifically, the default system.conf does not allow ownership unless specified otherwise in other included configuration files (commonly under /etc/dbus-1/system.d).
Additionally, if the name requested already exists–the request will not be granted until the owning process releases the name.
Our team has started enumerating services that run as root and listen to messages on the System Bus, performing both code reviews and dynamic analysis. We have reported two information leak issues as a result:
While these are interesting, their severity is low – an attacker can list files under directories that require high permissions to list files under. Then we started noticing interesting patterns in a systemd unit called networkd-dispatcher. The goal of networkd-dispatcher is to dispatch network status changes and optionally perform different scripts based on the new status. Interestingly, it runs on boot as root:
Upon examination of the networkd-dispatcher source code, we noticed an interesting flow:
Step 5 has multiple security issues:
Let us assume an adversary has a malicious D-Bus component that can send an arbitrary signal. An attacker can therefore do the following:
Note that to win the TOCTOU race condition with high probability, we plant many files that can potentially run. Our experiments show three attempts were enough to win the TOCTOU race condition.
Since we do not wish to run the exploit every time we want to run as root, the payload that we ended up implementing leaves a root backdoor as such:
The astute reader will notice that the entire exploit elevates privileges assuming our exploit code can own the “org.freedesktop.network1” bus name. While this sounds non-trivial, we have found several environments where this happens. Specifically:
The query we used can also be run by Microsoft Defender for Endpoint customers:
DeviceProcessEvents
| where Timestamp > ago(5d)
and AccountName == "systemd-network"
and isnotempty(InitiatingProcessAccountName)
and isnotempty(FileName)
| project DeviceId, FileName, FolderPath, ProcessCommandLine
We were therefore able to exploit these scenarios and implement our own exploit:
While capable of running any arbitrary script as root, our exploit copies /bin/sh to the /tmp directory, sets /tmp/sh as a Set-UID (SUID) executable, and then invokes “/tmp/sh -p”. Note that the “-p” flag is necessary to force the shell to not drop privileges.
Despite the evolving threat landscape regularly delivering new threats, techniques, and attack capabilities, adversaries continue to focus on identifying and taking advantage of unpatched vulnerabilities and misconfigurations as a vector to access systems, networks, and sensitive information for malicious purposes. This constant bombardment of attacks spanning a wide range of platforms, devices, and other domains emphasizes the need for a comprehensive and proactive vulnerability management approach that can further identify and mitigate even previously unknown exploits and issues.
Microsoft’s threat and vulnerability management capabilities help organizations monitor their overall security posture, providing real-time insights into risk with continuous vulnerability discovery, contextualized intelligent prioritization, and seamless one-click flaw remediation. Leveraging our research into the Nimbuspwn vulnerabilities to improve solutions, our threat and vulnerability management already covers CVE-2022-29799 and CVE-2022-29800 and indicates such vulnerable devices in the threat and vulnerability module in Microsoft Defender for Endpoint.
To address the specific vulnerabilities at play, Microsoft Defender for Endpoint’s endpoint detection and response (EDR) capabilities detect the directory traversal attack required to leverage Nimbuspwn. Additionally, the Microsoft Defender for Endpoint detection team has a generic detection for suspicious Set-UID process invocations, which detected our exploit without prior knowledge.
Defending against the evolving threat landscape requires the ability to protect and secure users’ computing experiences, be it a Windows or non-Windows device. Microsoft continuously enriches our protection technologies through robust research that protects users and organizations across all the major platforms every single day. This case displayed how the ability to coordinate such research via expert, cross-industry collaboration is vital to effectively mitigate issues, regardless of the vulnerable device or platform in use. By sharing our research and other forms of threat intelligence, we can continue to collaborate with the larger security community and strive to build better protection for all.
Jonathan Bar Or
Microsoft 365 Defender Research Team
Microsoft has been tracking activity related to the financially motivated threat actor Octo Tempest, whose evolving campaigns represent a growing concern for many organizations across multiple industries.
A set of memory corruption vulnerabilities in the ncurses library could have allowed attackers to chain the vulnerabilities to elevate privileges and run code in the targeted program's context or perform other malicious actions.
In a recent investigation by Microsoft Incident Response of a BlackByte 2.0 ransomware attack, we found that the threat actor progressed through the full attack chain, from initial access to impact, in less than five days, causing significant business disruption for the victim organization.
Microsoft has uncovered an attack leveraging custom and open-source tools to target internet-facing IoT devices and Linux-based systems. The attack involves deploying a patched version of OpenSSH on affected devices to allow root login and the hijack of SSH credentials.