Researchers from ETH Zurich have revealed details of Ahoi attacks, a novel method targeting confidential virtual machines (CVMs) in cloud environments. These attacks exploit vulnerabilities within hardware-based trusted execution environments, specifically AMD’s SEV-SNP and Intel’s TDX technologies, used to protect VMs in cloud platforms. Through two variations, Heckler and WeSee, malicious hypervisors inject interrupts to compromise CVM integrity, gain root access, and extract sensitive information such as TLS session keys.
Upon discovering the vulnerabilities, the ETH Zurich researchers notified major stakeholders, including Intel, AMD, AWS, Microsoft, and Google, prompting the release of Linux kernel patches and mitigations. Despite efforts to address the issue, cloud vendors exhibit varied impacts, with AWS planning to resolve kernel issues affecting Amazon Linux in a future release. However, Microsoft’s Azure claims non-impact, while Google’s stance remains undisclosed, underscoring the ongoing need for heightened vigilance and security measures in cloud environments.
The Ahoi attacks pose a significant threat to cloud security, as they exploit fundamental vulnerabilities within trusted execution environments designed to protect CVMs from unauthorized access. With CVE identifiers assigned to these vulnerabilities, the research underscores the critical importance of timely detection, notification, and remediation efforts in mitigating potential risks to cloud infrastructure and sensitive data.
