Out-of-bounds read in TensorFlow - CVE-2021-29547
Published: May 13, 2021 / Updated: May 4, 2026
Vulnerability identifier: #VU129345
CSH Severity: Medium
CVSSv4.0: CVSS:4.0/AV:N/AC:L/AT:N/PR:N/UI:N/VC:N/VI:N/VA:L/SC:N/SI:N/SA:N/E:U/U:Green
CVE-ID: CVE-2021-29547
CWE-ID: CWE-125
Exploitation vector: Remote access
Exploit availability:
No public exploit available
Vulnerable software:
TensorFlow
TensorFlow
Software vendor:
TensorFlow
TensorFlow
Description
The vulnerability allows a remote attacker to cause a denial of service.
The vulnerability exists due to an out-of-bounds read in tf.raw_ops.QuantizedBatchNormWithGlobalNormalization when processing empty input tensors. A remote attacker can supply crafted empty inputs to cause a denial of service.
The issue is triggered when one of the scalar min or max input tensors is empty and the implementation accesses element 0 of the flattened buffer.
Remediation
Install security update from vendor's website.