HomeWinBuzzer NewsResearchers Expose Security Flaws in ARM Memory Tagging Extension

Researchers Expose Security Flaws in ARM Memory Tagging Extension

Researchers have found critical flaws in ARM's MTE security system. These "TikTag" vulnerabilities allow attackers to bypass MTE protections,


A collaborative research team from Seoul National University, Samsung Research, and Georgia Institute of Technology has identified critical vulnerabilities in ARM's Memory Tagging Extension (MTE). These vulnerabilities, exploitable via speculative execution attacks, allow for the bypassing of MTE protections using newly discovered mechanisms, termed TikTag-v1 and TikTag-v2. These gadgets leak MTE tags from any memory address, thus exposing the system to potential exploitation.

Background on ARM MTE

ARM introduced the Memory Tagging Extension (MTE) in 2018 to improve memory safety by associating 16-byte memory blocks with unique four-bit metadata tags. This security measure aims to detect and prevent memory corruption by ensuring that memory pointers contain matching tags. Any tag mismatch leads to an error, which serves as a safeguard against errors and malicious exploits. MTE technology has been adopted in devices like Google's Pixel 8 series and is incorporated within the Linux kernel as a defense against prevalent memory issues, including buffer overflows and use-after-free vulnerabilities.

Speculative Execution Attacks Explained

Speculative execution attacks take advantage of the CPU's behavior to potentially access confidential information. Spectre and Meltdown are prominent examples of such attacks. These methods leverage speculative execution to load and execute prohibited data paths, leading to potential data leakage via cache state analysis. Attackers gain access to sensitive information by observing cache behaviors and measuring access times to detect hidden data values.

The researchers demonstrated that TikTag-v1 and TikTag-v2 could achieve a tag leakage success rate of about 95% within four seconds. These gadgets manipulate tag check results, influencing CPU data prefetching and blocking store-and-load forwarding when a tag check fault happens. Such capabilities severely compromise the integrity of any MTE-based security mechanisms employed in environments like Chrome and the Linux kernel.

Real-World Implications

In practical demonstrations, researchers executed proof-of-concept attacks on Google Chrome and the Linux kernel, leveraging TikTag gadgets. For Chrome, they exploited the V8 JavaScript engine to extract MTE tags from the memory. Similarly, within the Linux kernel, they identified a TikTag gadget capable of revealing tag check results from kernel memory, significantly weakening systems that depend on MTE for security.

The research team proposed several mitigation approaches to address these new vulnerabilities. These included the introduction of speculation barriers to halt speculative tag leakage and restricting TikTag gadget implementations within Chromium and Linux kernel codebases. ARM has acknowledged the vulnerabilities and has recommended similar mitigations. However, Google's Chrome team decided against adjustments in the V8 engine, citing that it does not guarantee the confidentiality of memory and MTE tags. The Android Security Team has responded by addressing the discovered vulnerabilities in Pixel 8 devices and has issued a bug bounty for the researchers' findings.

Luke Jones
Luke Jones
Luke has been writing about all things tech for more than five years. He is following Microsoft closely to bring you the latest news about Windows, Office, Azure, Skype, HoloLens and all the rest of their products.