Recent discoveries by computer scientists at ETH Zurich have raised concerns about the security of semiconductor technologies, particularly in relation to confidential computing, as reported in TechXplore.
What is confidential computing?
Confidential computing is a method used by businesses and government organizations to process sensitive data securely in shared cloud computing environments.
However, researchers have identified vulnerabilities in the latest chip technologies that could potentially compromise data security.
Two attack scenarios
The vulnerabilities were discovered using two attack scenarios devised by the Secure & Trustworthy Systems Group at ETH Zurich.
These scenarios exploit the interrupt mechanism, which temporarily interrupts regular processing to prioritize other computing tasks. The research team, led by Professor Shweta Shinde, discovered that certain interrupts could be manipulated to gain unauthorized access to sensitive data stored on server hardware from major chip manufacturers such as AMD and Intel.
The vulnerabilities were discovered during an investigation into sensitive computing technologies used by AMD and Intel processors. The researchers sought to understand the functionality of these processors as part of their work to develop an eavesdropping-proof smartphone based on confidential computing principles.
According to the research team, at the heart of confidential computing is the trusted execution environment (TEE), a hardware-based component that isolates applications during runtime.
This isolation ensures that sensitive data is protected from unauthorized access while it is being processed. However, vulnerabilities in the interrupt handling mechanisms of these processors pose a significant risk to data security.
One vulnerability identified by the research team involves the interaction between hypervisors and TEEs in public cloud environments. Hypervisors, which manage virtualized resources in cloud environments, are essential to ensure the flexibility, efficiency and security of cloud services.
However, they also introduce potential security risks, as they can be exploited to access data stored in the memory of other cloud users or by the cloud providers themselves.
“Ahoi attacks”
The researchers carried out what they called “Ahoi attacks” to exploit the hypervisor-TEE interface and send coordinated outages to the secure system. These attacks demonstrated that certain outages could bypass TEE security measures, allowing unauthorized access to sensitive data.
Another attack scenario, known as WeSee, exploits a mechanism introduced by AMD to facilitate communication between the TEE and the hypervisor. This mechanism can be manipulated to extract sensitive data from the secure system or execute external programs.
The vulnerabilities identified by the research team highlight the challenges of ensuring data security in sensitive computing environments. While chipmakers have taken steps to address these vulnerabilities, the risks associated with interrupt-based attacks remain a concern for enterprises and government organizations, according to the team.
For Professor Shinde and her research group, uncovering these vulnerabilities is part of a broader effort to ensure user control over data and applications on smartphones.
By developing a TEE that supports unsupervised operation of applications not managed by device manufacturers, your goal is to improve user privacy and security in mobile computing environments.
