Microarchitecture Vulnerabilities: Uncovering The Root Cause Weaknesses

Understanding the characteristics of transient execution vulnerabilities helps prevent future ones.


In early 2018, the tech industry was shocked by the discovery of hardware microarchitecture vulnerabilities that bypassed decades of work put into software and application security. Meltdown and Spectre exploited performance features in modern application processors to leak sensitive information about victim programs to an adversary. This leakage occurs through the hardware itself, meaning that malicious software can extract secret information from users even if software protections are in place because the leakages happen below the view of software in hardware. Since these so-called transient execution vulnerabilities were first publicly disclosed, dozens of variants have been identified that all share a set of common root cause weaknesses, but the specifics of that commonality were not well understood broadly by the security community.

In early 2020, Intel Corporation, MITRE, Cycuity, and others set off to establish a set of common weaknesses for hardware to enable a more proactive approach to hardware security to reduce the risk of a hardware vulnerability in the future. The initial set of weaknesses, in the form of Common Weakness Enumerations (CWE), were broad and covered weaknesses beyond just transient execution vulnerabilities like Meltdown and Spectre. While this initial set of CWEs was extremely effective at covering the root causes across the entire hardware vulnerability landscape, the precise and specific coverage of transient execution vulnerabilities was still lacking. This was primarily because of the sheer complexity, volume, and cleverness of each of these vulnerabilities.

In the fall of 2022, technical leads from AMD, Arm, Intel (special kudos to Intel for initiating and leading the effort), Cycuity, and Riscure came together to dig into the details of publicly disclosed transient execution vulnerabilities to really understand their root cause and come up with a set of precise, yet comprehensive, root cause weaknesses expressed as CWEs to help the industry not only understand the root cause for these microarchitecture vulnerabilities but to help prevent future, unknown vulnerabilities from being discovered. The recent announcement of the four transient execution weaknesses was a result of this collaborative effort over the last year.

CWEs for microarchitecture vulnerabilities

To come up with these root cause weaknesses, we researched every known publicly disclosed microarchitecture vulnerability (Common Vulnerabilities and Exposures [CVEs]) to understand the exact characteristics of the vulnerabilities and what the root causes were. As a result of this, the following common weaknesses were discovered, with a brief summary provided in layman terms from my perspective:

CWE-1421: Exposure of Sensitive Information in Shared Microarchitectural Structures during Transient Execution

  • Potentially leaky microarchitectural resources are shared with an adversary. For example, sharing a CPU cache between victim and attacker programs has shown to result in timing side channels that can leak secrets about the victim.

CWE-1422: Exposure of Sensitive Information caused by Incorrect Data Forwarding during Transient Execution

  • The forwarding or “flow” of information within the microarchitecture can result in security violations. Often various events (speculation, page faults, etc.) will cause data to be incorrectly forwarded from one location of the processor to another (often to a leaky microarchitecture resource like the one listed in CWE-1421)

CWE-1423: Exposure of Sensitive Information caused by Shared Microarchitectural Predictor State that Influences Transient Execution

  • An attacker being able to affect or “poison” a microarchitecture predictor used within the processor. For example, branch prediction is commonly used to increase performance to speculatively fetch instructions based on the expected outcome of a branch in a program. If an adversary is able to affect the branch prediction itself, they can cause the victim to execute code in branches of their choosing.

CWE-1420: Exposure of Sensitive Information during Transient Execution

  • A general transient execution weakness if one of the other weaknesses above do not quite fit the need.

Within each of the CWEs listed above, you can find details about observed examples, or vulnerabilities, which are a result of these weaknesses. Some vulnerabilities, Spectre-V1, for example, requires the presence of CWE-1421, CWE-1422, and CWE-1423. While others, like Meltdown, only require CWE-1422 and CWE-1423.

Since detecting these weaknesses can be a daunting task, each of the CWEs outline a set of detection methods. One detection method that is highlighted in each CWE entry is the use of information flow to track the flow of information in the microarchitecture to ensure data is being handled securely. Information flow can be used for each of the CWEs as follows:

  • CWE-1421: information flow analysis can be used to ensure that secrets never end up in a shared microarchitectural resource.
  • CWE-1422: ensure that secret information is never improperly forwarded within the microarchitecture.
  • CWE-1423: ensure that an attacker can never affect or modify the predictor state in a way that is observable by the victim. In other words, information from the attacker should not flow to the predictor if that information can affect the integrity of the predictor for the victim.

Our Radix products use information flow at their core and we have already shown success in demonstrating Radix’s ability to detect Meltdown and Spectre. We look forward to continuing to work with the industry and our customers and partners to further advance the state of hardware security and reduce the risk of vulnerabilities being discovered in the future.

Leave a Reply

(Note: This name will be displayed publicly)