NVIDIA TensorRT contains a vulnerability where an attacker could cause an out-of-bounds write. A successful exploit of this vulnerability might lead to data tampering.
CVSS Details
CVSS Score
8.2
Severity
HIGH
CVSS Vector
CVSS:3.1/AV:N/AC:L/PR:N/UI:N/S:U/C:N/I:H/A:L
Configurations (Affected Products)
No configuration data available.
NVIDIA TensorRT (具体受影响版本未在提供的信息中明确列出，请参考官方公告)
PoC / Exploit Code
⚠ For Security Research Only
The following code is for security research and authorized testing only.
python
import sys
# Conceptual Proof of Concept for CVE-2026-24188
# This script simulates the logic leading to an Out-of-Bounds Write
# In a real scenario, this would be triggered by a specific input to TensorRT.
def simulate_tensorrt_parsing(data_stream):
# Simulated internal buffer in TensorRT
internal_buffer = [0] * 256
try:
# Parsing input index from attacker controlled stream
offset = int(data_stream.get('offset', 0))
payload = data_stream.get('data', 0x00)
# VULNERABILITY: Lack of bounds checking on 'offset'
# If offset is larger than 255, this writes out of bounds.
print(f"[*] Writing payload {hex(payload)} to buffer offset {offset}")
internal_buffer[offset] = payload
print("[+] Write successful. Memory integrity potentially compromised.")
except IndexError:
# Python handles this, but C++ (TensorRT's language) would corrupt memory
print("[-] Runtime Error: Index out of bounds (Handled by Python interpreter)")
print("[!] In a vulnerable C++ environment, this would result in OOB Write.")
# Attacker controlled payload
malicious_input = {
'offset': 300, # Exceeds buffer size of 256
'data': 0xDEADBEEF
}
print("--- CVE-2026-24188 PoC Simulation ---")
simulate_tensorrt_parsing(malicious_input)