There has been an announcement of a new technique for bypassing key security protections used in AMD chips to gain access to the clients of those services. Researchers believe that hackers will be able to spy on clients through physical access to cloud computing environments. Known as the "badRAM" security flaw, it has been described as a $10 hack that undermines the trust that the cloud has in it. 

This vulnerability was announced on Tuesday. Like other branded vulnerabilities, this vulnerability is being disclosed on a website with a logo and will be explained in a paper to be presented at next May's IEEE Symposium on Security and Privacy 2025. 

There is an increasing use of encryption in today's computers to protect sensitive data in their DRAM, especially in shared cloud environments with multiple data breaches and insider threats, which are commonplace. The Secure Encrypted Virtualization (SEV) technology of AMD enables users to protect privacy and trust in cloud computing by encrypting the memory of virtual machines (VMs) and isolating them from advanced attackers, including those who compromise critical infrastructure like the virtual machine manager and firmware, which is a cutting-edge technology. 

According to researchers, AMD's Secure Encrypted Virtualization (SEV) program, which protects processor memory from prying eyes in virtual machine (VM) environments, is capable of being tricked into letting someone access the contents of its encrypted memory using a test rig which costs less than $10 and does not require additional hardware. It is important to note that AMD is among the first companies to leverage the capabilities of chipset architecture to improve processor performance, efficiency, and flexibility. 

It has been instrumental in extending and building upon Moore's Law performance gains and extending them further. As a result of the firm's research, performance gains under Moore's Law have been extended and built upon, and the company announced in 2018 that the first processor would have a chipset-based x86 CPU design that was available. Researchers at the University of Lübeck, KU Leven, and the University of Birmingham have proposed a conceptually easy and cheap attack called “BadRAM”. 

It consists of a rogue memory module used to trick the CPU into believing that it has more memory than it does. Using this rogue memory module, you get it to write its supposedly secret memory contents into a "ghost" space that is supposed to contain the hidden memory contents. In order to accomplish this task, researchers used a test rig anyone could afford to buy, composed of a Raspberry Pi Pico, which costs a couple of dollars, and a DIMM socket for DDR4/5 RAM modules. 

The first thing they did was manipulate the serial presence detection (SPD) chip within the memory module so that it would misreport the amount of memory onboard when the device was booted up – the “BadRAM” attack. Using reverse engineering techniques to locate these memory aliases, they had access to memory contents by bypassing the system's trusted execution environment (TEE), as this created two physical addresses referencing the same DRAM location. 

According to the CVE description, the issue results from improper input validation of DIM SPD metadata, which could potentially allow an attacker with certain access levels to overwrite guest memory, as the issue is described as a result of improper input validation. It has been deemed a medium severity threat on the CVSS, receiving a 5.3 rating owing to the high level of access that a potential attacker would need to engage to successfully exploit the problem. 

According to AMD, the issue may be a memory implementation issue rather than a product vulnerability, and the barriers to committing the attack are a lot higher than they would be if it were a software product vulnerability. AMD was informed of the vulnerability by the researchers in February, which has been dubbed CVE-2024-21944, as well as relates specifically to the company’s third and fourth-generation EPYC enterprise processors. According to AMD’s advisory, the recommendation is to use memory modules that lock SPD and to follow physical security best practices. 

A firmware update has also been issued, although each OEM's BIOS is different, according to AMD. As the company has stated on several occasions, it will make mitigations more prominent in the system; there is specific information on the condition of a Host OS/Hypervisor, and there is also information available on the condition of a Virtual Machine (Guest) to indicate that mitigation has been applied.

The AMD company has provided an in-depth explanation of the types of access an attacker would need to exploit this issue in a statement given to ITPro, advising clients to follow some mitigation strategies to prevent the problem from becoming a problem. The badRAM website states that this kind of tampering may occur in several ways — either through corrupt or hostile employees at cloud providers or by law enforcement officers with physical access to the computer. 

In addition, the badRAM bug may also be exploited remotely, although the AMD memory modules are not included in this process. All manufacturers, however, that fail to lock the SPD chip in their memory modules, will be at risk of being able to modify their modules after boot as a result of operating system software, and thus by remote hackers who can control them remotely. 

According to Recorded Future News, Oswald has said that there has been no evidence of this vulnerability being exploited in the wild. However, the team discovered that Intel chips already had mitigations against badRAM attacks. They could not test Arm's modules because they were unavailable commercially. An international consortium of experts led by researchers from KU Leuven in Belgium; the University of Luebeck in Germany; and the University of Birmingham in the United Kingdom conducted the research.