Model-execution hang via non-reentrant recursive function lock
Published Nov 5, 2021 · Updated Aug 4, 2024
Denial of service in Google TensorFlow before 2.4.4, 2.5.0 through 2.5.1, and 2.6.0 allows context-dependent attackers to hang model execution. Function wrappers in tensorflow/python/eager/def_function.py and function.py reacquire a non-reentrant threading.Lock when decorated functions recurse. A user must load an attacker-influenced model containing mutually recursive decorated functions and invoke one, leaving the executing process hung.
Summary
What happened
Denial of service in Google TensorFlow before 2.4.4, 2.5.0 through 2.5.1, and 2.6.0 allows context-dependent attackers to hang model execution. Function wrappers in tensorflow/python/eager/def_function.py and function.py reacquire a non-reentrant threading.Lock when decorated functions recurse. A user must load an attacker-influenced model containing mutually recursive decorated functions and invoke one, leaving the executing process hung.
The record
- CVE
- CVE-2021-41213
- Published
- Nov 5, 2021
- Updated
- Aug 4, 2024
- Vendor
- Product
- TensorFlow
- Classifications
- CWE-667, CWE-662, T1499.004
- Attack vector
- local
- Privileges
- Unavailable
Timeline
How it unfolded
- Nov 5, 2021CVE publishedPublication date reported by the CVE source.
- Aug 4, 2024Record updatedLatest update available in the CVE record.
Exploitability
Present is not the same as exploitable
Compare your product and version with the public record. A matching version still requires validation against your environment.
Is a vulnerable build present?
Compare these published version ranges with your installed build and any vendor patches.
- Affected versionversion=< 2.4.4
- Affected versionversion=>= 2.5.0, < 2.5.2
- Affected versionversion=>= 2.6.0, < 2.6.1
What conditions does exploitation require?
What is affected?
Published CVSS scores
CVSS describes severity. EPSS estimates exploitation probability.
Attacks
What attackers are doing with it
Daily unique IPs observed by Shadowserver honeypots for known exploited vulnerabilities (KEVs). Missing observations do not establish an absence of attacks.
Weakness, pattern, technique
Public exploit references
- Mutually recursive tf.function deadlock demonstrationproof of concept · demonstrated
Labels summarize the accepted research assessment. They do not indicate a test against your environment.
Technologies
Your stack
See the directory against your own environment.
Your stack
Check the software in your environment
Book a demo to see how Hinoki identifies affected software and validates exploitability in your environment.
Book a demo