NVIDIA TRT-LLM Vulnerability: Unsafe Deserialization Leads to RCE
The National Vulnerability Database has disclosed CVE-2026-24163, a high-severity vulnerability in NVIDIA TRT-LLM’s RPC testing component. This flaw, categorized as CWE-502 (Deserialization of Untrusted Data), allows an attacker to trigger unsafe deserialization. The CVSSv3.1 score of 7.5 (High) reflects the critical impact potential.
A successful exploit of this vulnerability could lead to remote code execution, denial of service, data tampering, and information disclosure. The attack vector is local, but the high privileges required (PR:H) and high attack complexity (AC:H) slightly temper the overall risk. Nonetheless, any RCE vector within AI/ML infrastructure demands immediate attention.
For defenders, this means scrutinizing environments where NVIDIA TRT-LLM is deployed. Unsafe deserialization is a well-understood attack primitive, and its presence in a crucial component like TRT-LLM for large language models is concerning. Patching is paramount, but a deeper look at RPC security and input validation practices is also warranted.
What This Means For You
- If your organization utilizes NVIDIA TRT-LLM, you must immediately assess your exposure to CVE-2026-24163. Prioritize patching this vulnerability and audit your RPC testing configurations for any potential unauthorized access or unusual activity. This isn't just about patching; it's about understanding the deserialization risks in your AI pipeline.
Related ATT&CK Techniques
🛡️ Detection Rules
3 rules · 6 SIEM formats3 detection rules auto-generated for this incident, mapped to MITRE ATT&CK. Sigma YAML is free — export to any SIEM format via the Intel Bot.
CVE-2026-24163 - TRT-LLM Unsafe Deserialization via RPC Testing
title: CVE-2026-24163 - TRT-LLM Unsafe Deserialization via RPC Testing
id: scw-2026-05-20-ai-1
status: experimental
level: high
description: |
Detects the execution of NVIDIA TRT-LLM's RPC testing utility, which is a potential vector for the unsafe deserialization vulnerability (CVE-2026-24163). This rule specifically targets the known testing component that could be leveraged by an attacker.
author: SCW Feed Engine (AI-generated)
date: 2026-05-20
references:
- https://shimiscyberworld.com/posts/nvd-CVE-2026-24163/
tags:
- attack.initial_access
- attack.t1190
logsource:
category: process_creation
detection:
selection:
Image|contains:
- 'trt-llm'
CommandLine|contains:
- 'rpc_testing'
condition: selection
falsepositives:
- Legitimate administrative activity
Source: Shimi's Cyber World · License & reuse
Indicators of Compromise
| ID | Type | Indicator |
|---|---|---|
| CVE-2026-24163 | Deserialization | NVIDIA TRT-LLM for any platform |
| CVE-2026-24163 | RCE | NVIDIA TRT-LLM RPC testing unsafe deserialization |
| CVE-2026-24163 | DoS | NVIDIA TRT-LLM RPC testing unsafe deserialization |
| CVE-2026-24163 | Information Disclosure | NVIDIA TRT-LLM RPC testing unsafe deserialization |
Written by SCW Vulnerability Desk
Source & Attribution
| Source Platform | NVD |
| Channel | National Vulnerability Database |
| Published | May 20, 2026 at 07:16 UTC |
This content was AI-rewritten and enriched by Shimi's Cyber World based on the original source. All intellectual property rights remain with the original author.
Believe this infringes your rights? Submit a takedown request.
Related coverage
CVE-2026-20240 — Denial of Service
CVE-2026-20240 — In Splunk Enterprise versions below 10.2.2, 10.0.5, 9.4.11, and 9.3.12, and Splunk Cloud Platform versions below 10.4.2603.1, 10.3.2512.9, 10.2.2510.11, 10.1.2507.21, 10.0.2503.13, and 9.3.2411.129,...
Splunk Enterprise, Cloud Vulnerability Exposes Session Cookies, Sensitive Data
CVE-2026-20239 — In Splunk Enterprise versions below 10.2.2 and 10.0.5, and Splunk Cloud Platform versions below 10.3.2512.8, 10.2.2510.11, 10.1.2507.21, and 10.0.2503.13, a user with a...
CVE-2026-20238 — In Splunk AI Toolkit versions below 5.7.3, a low-privileged
CVE-2026-20238 — In Splunk AI Toolkit versions below 5.7.3, a low-privileged user that does not hold the 'admin' or 'power' roles could access confidential data...