CVE-2023-31024

9.0 CRITICAL

📋 TL;DR

This vulnerability in NVIDIA DGX A100 BMC's host KVM daemon allows unauthenticated attackers to cause stack memory corruption via specially crafted network packets. Successful exploitation could lead to arbitrary code execution, denial of service, information disclosure, or data tampering. Only NVIDIA DGX A100 systems with vulnerable BMC firmware are affected.

💻 Affected Systems

Products:
  • NVIDIA DGX A100
Versions: BMC firmware versions prior to 00.19.07
Operating Systems: BMC firmware (not host OS)
Default Config Vulnerable: ⚠️ Yes
Notes: Only affects the Baseboard Management Controller (BMC) component, not the main host operating system.

📦 What is this software?

⚠️ Risk & Real-World Impact

🔴

Worst Case

Complete system compromise with attacker gaining full control over the BMC, potentially leading to persistent access, data theft, and disruption of the entire DGX A100 system.

🟠

Likely Case

Denial of service causing BMC unavailability, potentially disrupting system management capabilities and requiring physical intervention.

🟢

If Mitigated

Limited impact if network segmentation prevents direct BMC access from untrusted networks.

🌐 Internet-Facing: HIGH if BMC interfaces are exposed to the internet, as exploitation requires no authentication.
🏢 Internal Only: MEDIUM if internal network segmentation exists, but still concerning due to unauthenticated nature.

🎯 Exploit Status

Public PoC: ✅ No
Weaponized: UNKNOWN
Unauthenticated Exploit: ⚠️ Yes
Complexity: MEDIUM

Exploitation requires network access to BMC interfaces and crafting specific packets, but no authentication is needed.

🛠️ Fix & Mitigation

✅ Official Fix

Patch Version: BMC firmware version 00.19.07 or later

Vendor Advisory: https://nvidia.custhelp.com/app/answers/detail/a_id/5510

Restart Required: Yes

Instructions:

1. Download BMC firmware update from NVIDIA support portal. 2. Follow NVIDIA's DGX A100 BMC firmware update procedure. 3. Reboot the BMC after update completion.

🔧 Temporary Workarounds

Network Segmentation

all

Isolate BMC management interfaces from untrusted networks and restrict access to authorized management systems only.

Access Control Lists

all

Implement network ACLs to restrict traffic to BMC ports (typically 623/UDP, 664/TCP for IPMI) to trusted IP addresses only.

🧯 If You Can't Patch

  • Implement strict network segmentation to isolate BMC interfaces from all untrusted networks
  • Deploy intrusion detection systems to monitor for suspicious BMC network traffic patterns

🔍 How to Verify

Check if Vulnerable:

Check BMC firmware version via IPMI tool: 'ipmitool mc info' or through NVIDIA's management interface.

Check Version:

ipmitool mc info | grep 'Firmware Revision'

Verify Fix Applied:

Confirm BMC firmware version is 00.19.07 or later using the same version check methods.

📡 Detection & Monitoring

Log Indicators:

  • Unusual BMC authentication failures
  • BMC service crashes or restarts
  • Suspicious network connections to BMC ports

Network Indicators:

  • Unusual traffic patterns to BMC IPMI ports (623/UDP, 664/TCP)
  • Malformed packets targeting BMC services

SIEM Query:

source_ip IN (untrusted_networks) AND dest_port IN (623, 664) AND protocol IN (UDP, TCP)

🔗 References

📤 Share & Export