Intel has effectively resolved a significant security vulnerability known as “Downfall” that was discovered by Daniel Moghimi from the University of California San Diego. This vulnerability was described as a crucial flaw present in modern processors numbering in the billions. It affects Intel processors released between 2015 and 2019, specifically those from the 6th Gen Skylake to the 11th Gen Tiger Lake. Intel has provided a detailed list of the affected chip models and has chosen to refer to this vulnerability as “Collect Data Sampling” (GDS) rather than “Downfall.”
The impact of the vulnerability is significant, as it allows attackers to target valuable credentials like passwords and encryption keys. Interestingly, the flaw only requires the attacker and victim to share the same physical CPU core, which might seem improbable but is highly likely in the context of multitasking, multithreading, servers, and cloud computing.
Intel’s response to the issue includes a fix for the vulnerability. However, this fix does come with a notable performance impact on certain types of workloads. Scientific and visualization engineering tasks are expected to experience the most significant performance reduction, with the overhead of the mitigation potentially reaching up to 50% depending on the workload.
To address these concerns, Intel has implemented a mitigation mechanism that can be disabled, although it is activated by default. This introduction of an opt-out feature raises questions about which Intel servers will remain vulnerable, as the decision to apply the fix is left to the server owner.
Intel has responded to Moghimi’s findings by stating that his research was conducted in a controlled research environment, and the execution of such an attack “would be exceedingly difficult to execute beyond such controlled circumstances.”
