BadRAM: $10 Hack That Breaks AMD’s Cloud Encryption Technology

Researchers have unveiled a critical vulnerability in AMD's Secure Encrypted Virtualization (SEV) technology, which secures sensitive data in cloud environments. By exploiting manipulated DRAM modules in an attack termed “BadRAM,” attackers can bypass SEV protections, potentially undermining the trust in virtual machines (VMs) hosted in shared cloud infrastructure—all with a setup costing less than $10.

The vulnerability was detailed by researchers from institutions including KU Leuven, the University of Lübeck, and the University of Birmingham. Led by Jesse De Meulemeester, the team developed a method to modify the embedded Serial Presence Detect (SPD) chips on DRAM modules. These chips report memory configuration data to the processor during boot. By tampering with this data, the processor can be tricked into accessing false memory mappings, enabling attackers to execute aliasing attacks.

BadRAM mechanism

• Tampered Memory Configuration: The SPD chip is modified to misrepresent memory properties, tricking the CPU into accepting ghost addresses that alias to legitimate ones.
• Memory Aliases Located: Practical tools identify these aliases in minutes.
• Access Control Bypass: Using these aliases, attackers gain unauthorized access to sensitive data, compromising encrypted memory and injecting backdoors.

Impact on AMD SEV technology

AMD's SEV protects VMs through encryption and access controls, isolating them even from compromised hypervisors. The latest iteration, SEV-SNP, enhances integrity with reverse mapping tables (RMPs). However, researchers demonstrated how BadRAM allows attackers to:

1. Manipulate memory configurations, enabling ciphertext replay or tampering.

2. Bypass SEV-SNP's attestation feature, crucial for validating the integrity of VMs.

This end-to-end attack undermines the trust model of AMD's SEV ecosystem by enabling remote attackers to alter VMs undetectably.

As cloud environments rely on memory encryption to safeguard tenant data, this vulnerability highlights a systemic risk. With modern processors from AMD and Intel emphasizing Trusted Execution Environments (TEEs) like SEV, TDX, and SGX, even these advanced mechanisms are challenged by low-cost hardware exploits.

Mitigation recommendations

AMD has issued firmware updates to validate DRAM configurations during boot and mitigate BadRAM attacks. The researchers emphasize broader defenses, including strengthening SPD protection by introducing non-modifiable configurations, using comprehensive memory validation during initialization, and employing integrity-protected and freshness-aware memory encryption systems.

BadRAM attacks typically require brief physical access to the Serial Presence Detect (SPD) chip on DRAM modules to modify their contents, making such exploits more challenging than purely remote attacks. This physical access requirement could mitigate risks in many scenarios, especially for consumer systems with controlled hardware environments.

However, in shared cloud settings, insider threats — such as employees or local law enforcement — can easily leverage physical access to deploy these attacks. Moreover, several DRAM manufacturers fail to secure the SPD chip against software modification. This oversight allows attackers with root privileges to manipulate the SPD via the operating system or compromised BIOS, potentially enabling BadRAM attacks without physical access.