A New Security Threat to Public Clouds From Old Vulnerabilities
Vulnerabilities found in processors from a range of vendors, which caused a lot of angst in 2018 but are now considered to be relatively harmless, could pose a threat to public cloud systems, according to new research.
In a presentation at the WHY2025 hacker conference in the Netherlands this week, anonymous researchers said they were able to leak private information from public clouds by combining the Spectre transient execution vulnerability and L1TF, another chip security flaw discovered seven years ago.
Spectre and another transient execution flaw, dubbed Meltdown, were disclosed in 2018, raising worries that instructions executed by the CPUs could give attackers access to sensitive data found in files cached from disk, cryptographic keys and whatever else was stored in memory.
Both threats were considered to be fixed through software patches and deemed impractical for threat actors to use in attacks, according to a summary of the presentation. The mitigations put in place by the chip makers made it more difficult to exploit the vulnerabilities and the fact that bad actors would need a form of remote code execution (RCE) to set off the necessary CPU instructions, they likely would use the RCE for other attack methods.
Threat to Clouds, Not ‘Regular Users’
There isn’t much of a threat for “regular users,” they wrote. However, the story is different for public cloud providers, as they wrote in a 15-page research paper.
“In particular, given that today’s clouds have large fleets of older CPUs that lack comprehensive, in-silicon fixes to a variety of transient execution vulnerabilities, the question arises whether sufficient software-based defenses have been deployed to stop realistic attacks – especially those using older, supposedly mitigated vulnerabilities,” they wrote.
The problem with public cloud providers is that “their business model is to provide remote code execution as a service, and to rent out shared hardware resources as efficiently as possible,” they wrote in their summary.
Misplaced Confidence
Comprehensive mitigation of Spectre and other transient execution flaws is expensive, so chip makers instead used “spot” mitigations, which make the vulnerabilities too difficult to exploit. There have been no reports of realistic attacks using the flaws, which has led to what the researchers called a “lack of concern” over older vulnerabilities.
That’s created a false sense of security, they wrote.
“We show that the practice of mitigating vulnerabilities in isolation, without removing the root cause, leaves systems vulnerable,” the researchers wrote. “By combining such ‘mitigated’ (and by themselves harmless) vulnerabilities, attackers may still craft an end-to-end attack that is more than the sum of its parts.”
Running L1TF Reloaded
They ran their new attack method – dubbed “L1TF Reloaded” – on dedicated host systems in Amazon Web Services (AWS) and Google Cloud and also demonstrated their attack on a local system running KVM-based hypervisors, adding that the attack method could be used against other hypervisors.
They informed both cloud providers about their experiment and which systems they were targeting, and also kept them up-to-date on the progress. The name refers to another since-mitigated flaw from 2018, L1TF, in Intel chips.
Overall, the researchers were able to leak sensitive data from the hypervisor as well as a co-tenant on Google Cloud, such as the TLS key from a Nginx server running on a guest virtual machine. On AWS cloud instances, they were able to leak non-sensitive information from the host, but not data from guest VMs. AWS said in a blog post that it sponsored some of the researchers’ work and that the protections against these types of attacks in its AWS Nitro System hardware and software and the Nitro Hypervisor included more than the patches or “reactive mitigation.” It’s the architectural design of the Nitro System that ensured guest data couldn’t be leaked, according to the cloud giant.
AWS, Google Shore Up Protections
In a security bulletin, Google said it had immediately mitigated the Spectre threat – CVE-2018-3646 – in 2018 and that since then, “we have been researching the residual risk and working with the upstream Linux community to remediate this risk.”
“Recently we worked with security researchers from academia to evaluate the state of the art of CPU security mitigations, and potential attack techniques not considered back in 2018,” the IT giant wrote. “Google has applied fixes to the affected assets, including Google Cloud, to mitigate the issue.”
The researchers noted that both Google Cloud and AWS have applied patches to their hypervisors to protect against L1TF Reloaded and are planning future security improvements. In addition, Google awarded them more than $150,000, the most distributed from its Cloud Vulnerability Reward Program.
Trust is Crucial in the Cloud
They added that, given the wide use of public clouds, there needs to be trust in the security of those environments.
“As governments and enterprises around the world rely on proprietary public clouds for on-demand access to scalable computing, today’s clouds have become a foundational component of our digital infrastructure,” they wrote. “Since all multitenant cloud solutions imply that users may share physical resources with untrusted other tenants, they require implicit trust in the security guarantees of the underlying hardware and system software to keep data safe from adversarial cotenants.”
That said, “providing such guarantees is challenging, especially in the presence of the ever-expanding arsenal of transient execution and side-channel exploits,” they wrote.
