https://www.comsys.rwth-aachen.de/tags/vesitrust/VeSiTRUSTHugo Blox Builder (https://hugoblox.com)en-usWed, 19 Aug 2026 00:00:00 +0000https://www.comsys.rwth-aachen.de/media/logo.svghttps://www.comsys.rwth-aachen.de/tags/vesitrust/https://www.comsys.rwth-aachen.de/openings/2026/2026-08-security-cvm-obserability/Wed, 19 Aug 2026 00:00:00 +0000https://www.comsys.rwth-aachen.de/openings/2026/2026-08-security-cvm-obserability/<h2 id="background">Background</h2> <p>Confidential Virtual Machines (CVMs), such as AMD SEV-SNP VMs, are designed to protect guest memory and execution state even from a potentially untrusted hypervisor or cloud infrastructure provider. While these technologies significantly strengthen workload isolation, they also make traditional monitoring and debugging more challenging. Conventional observability often rely on the hypervisor or host infrastructure and may expose sensitive information outside the trusted boundary of the CVM. Existing systems such as <a href="https://github.com/coconut-svsm/svsm" target="_blank" rel="noopener">COCONUT-SVSM</a> introduce a trusted execution layer inside the CVM and are beginning to explore observability-related services. However, it remains unclear how telemetry can be collected and exported securely without reintroducing the information leakage that confidential computing is intended to prevent. This thesis explores how useful operational visibility can be preserved while maintaining the confidentiality and integrity guarantees of CVMs.</p> <h2 id="your-approach">Your Approach</h2> <p>The goal of this thesis is to investigate how confidential virtual machines can remain observable without weakening the security guarantees provided by technologies such as AMD SEV-SNP (or other VM-based TEEs). In particular, the thesis will study how operational information such as logs, metrics, and diagnostic data can be made available for monitoring, debugging, and performance analysis while minimizing unnecessary exposure to the untrusted host infrastructure. A central objective is to determine how much visibility can be provided to operators without reintroducing the same trust assumptions that confidential computing is designed to remove. Based on this investigation, you will design and implement a prototype that demonstrates a practical approach to confidential observability. The work will also examine the security and systems-level trade-offs introduced by different observability mechanisms and trust boundaries. Finally, the prototype will be evaluated in terms of its security properties, performance overhead, and usefulness for operating and debugging confidential workloads.</p> <h2 id="what-we-offer">What We Offer</h2> <ul> <li>Opportunity to work with low-level systems software and trusted computing technologies</li> <li>Access to state-of-the-art hardware and software platforms for confidential computing research</li> <li>Hands-on experience in designing and implementing secure systems</li> <li>A collaborative and supportive research environment</li> </ul> <h2 id="what-we-expect">What We Expect</h2> <ul> <li>Interest in systems security, trusted computing, and virtualization</li> <li>Strong programming skills, <strong>preferably in Rust and C</strong></li> <li><strong>Knowledge of Linux kernel</strong> and hypervisor internals is a plus</li> <li>Ability to work independently and regular (and effective) communication</li> </ul> <h2 id="contact-and-application">Contact and Application</h2> <p><strong>This topic is intended as a Master thesis only, given the scope and complexity of the topic.</strong> If you are interested in this topic area, please approach Hajeong. Please include a <strong>short motivation</strong> and a <strong>transcript of records</strong>. We are looking forward to your application!</p>