New StackWarp Attack Threatens Confidential VMs on AMD Processors

Why It Matters

StackWarp undermines the core security guarantees of AMD's SEV‑SNP, exposing cloud‑based confidential workloads to insider or supply‑chain attacks. The vulnerability forces cloud providers and enterprises to reassess trust models and accelerate patch deployment.

Key Takeaways

• StackWarp exploits stack engine synchronization failure
• Impacts AMD Zen 1‑5, compromising confidential VMs
• Allows host‑controlled VM to steal keys, passwords
• AMD rates severity low; patches released July 2025
• CVE‑2025‑29943 assigned; mitigation requires firmware updates

Pulse Analysis

The emergence of StackWarp highlights a broader trend where microarchitectural flaws bypass traditional isolation mechanisms. By targeting the stack pointer update logic in the CPU frontend, the attack sidesteps memory encryption provided by AMD's SEV‑SNP, proving that even encrypted VM memory cannot guarantee integrity when control flow can be redirected. This revelation forces security teams to look beyond encryption and consider the full execution stack, especially in multi‑tenant cloud environments where a compromised host can become a potent vector.

For cloud service providers, the practical implications are immediate. While the vulnerability requires privileged access to the host, insider threats or advanced persistent threats that gain such footholds can exfiltrate cryptographic material, hijack user sessions, and persist within the infrastructure. The demonstrated ability to reconstruct RSA‑2048 keys and bypass SSH authentication underscores the risk to critical workloads, including financial services and SaaS platforms that rely on confidential VMs for data protection. Providers must prioritize rapid firmware updates, enforce strict host‑access controls, and augment monitoring to detect anomalous stack manipulations.

AMD's response—classifying the issue as low severity and releasing patches months after discovery—raises questions about disclosure timelines and the adequacy of existing mitigation pathways. Enterprises should verify that their EPYC servers run the latest microcode and consider layered defenses such as runtime integrity monitoring and hardware‑rooted attestation. As hardware vulnerabilities continue to evolve, a proactive stance combining timely patch management, rigorous access policies, and diversified security controls will be essential to preserve the confidentiality and integrity of virtualized workloads.