New LVI Intel CPU Data Theft Vulnerability Requires Hardware Fix

A novel class of attack techniques against modern Intel processors can allow threat actors to inject malicious data into applications via transient-execution attacks and steal sensitive data according to researchers.

The vulnerability dubbed LVI (short for Load Value Injection) and tracked as CVE-2020-0551 was discovered and reported to Intel on April 4, 2019, by researchers at the Worcester Polytechnic Institute, imec-DistriNet/KU Leuven, Graz University of Technology, University of Michigan, University of Adelaide and Data61, in no particular order.

Bitdefender researchers also independently discovered one variant of attack in the LVI class (LVI-LFB) and reported it to Intel in February 2020.

LVI attacks let attackers change the normal execution of programs to steal data that is normally meant to be kept private within SGX enclaves. Sensitive information that can be stolen this way includes passwords, private keys of certificates, and more.

Even though the Intel Software Guard eXtensions (SGX) feature in modern Intel processors that enables apps to run within secure and isolated enclaves is not necessary to launch an LVI attack, its presence makes the attack a lot easier.

"While LVI attacks in non-SGX environments are generally much harder to mount, we consider none of the adversarial conditions for LVI to be unique to Intel SGX," the researchers explain.

New Spectre-type data injection vulnerability

"LVI turns previous data extraction attacks around, like Meltdown, Foreshadow, ZombieLoad, RIDL and Fallout, and defeats all existing mitigations," the researchers explain.

"Instead of directly leaking data from the victim to the attacker, we proceed in the opposite direction: we smuggle — 'inject' — the attacker's data through hidden processor buffers into a victim program and hijack transient execution to acquire sensitive information, such as the victim’s fingerprints or passwords."

In short, LVI attacks allow injecting arbitrary data (much like Spectre attacks) within the memory loaded by a targeted application under certain conditions, making it possible for an attacker to hijack the control and data flow until the app rolls back all operations after detecting the mistake.

The new vulnerability bypasses all transient-execution attack mitigations developed for Intel's processors so far, like Meltdown, Spectre, Foreshadow, ZombieLoad, RIDL, and Fallout.

To exploit LVI, attackers would have to go through the following four steps:

Also, LVI is a lot harder to mitigate than previous Meltdown-type attacks because it needs expensive software patches that could potentially make Intel SGX enclave computations between two and 19 times slower.

Modern Intel processors affected, mitigations available

LVI affects Intel Core-family processors from Skylake onwards with SGX support and a list with all affected CPUs is provided by Intel here.

Icelake Core-family processors aren't affected by LVI, while Meltdown-resistant processors are "only potentially vulnerable to LVI-zero-data (aka loads exhibiting zero injection behavior only)."

Short term solutions for mitigating LVI have to be implemented to protect already deployed systems from potential attacks targeting this flaw.

"LVI necessitates compiler patches to insert explicit lfence speculation barriers which serialize the processor pipeline after potentially every vulnerable load instruction," the researchers say.

"Additionally and even worse, due to implicit loads, certain instructions have to be blacklisted, including the ubiquitous x86 ret instruction."

Even though software workarounds can be implemented, the root cause behind LVI cannot be fixed with software changes which means that new CPUs from affected processor families will need to come with hardware fixes.

"This is not a trivial attack to execute against a target, as several prerequisites have to be met," Bitdefender director of threat research Bogdan Botezatu told BleepingComputer. "This is not an average, run-of-the-mill malware attack that one would use against home users for instance."

"This is something that a determined threat actor, such as a hostile government-sponsored entity or a corporate espionage group would use against a high-profile target to leak mission-critical data from a vulnerable infrastructure.

"Although difficult to orchestrate, this type of attack would be impossible to detect and block by existing security solutions or other intrusion detection systems and would leave no forensic evidence behind."

Researchers have identified a new mechanism referred to as Load Value Injection (LVI). Due to the numerous complex requirements that must be satisfied to successfully carry out, Intel does not believe LVI is a practical method in real world environments where the OS and VMM are trusted. New mitigation guidance and tools for LVI are available now and work in conjunction with previously released mitigations to substantively reduce the overall attack surface. We thank the researchers who worked with us, and our industry partners for their contributions on the coordinated disclosure of this issue. - Intel

An academic research paper including more technical information regarding LVI is available here in PDF format and it will be presented in May 2020 at the 41st IEEE Symposium on Security and Privacy (IEEE S&P'20).

Proof of concept code detailing LVI attack applications is available on GitHub and Intel has also published a white paper here.

A video presenting demos of two LVI (Load Value Injection) proof of concept attacks is embedded below.

---

Update: Added Intel's statement.