Security readout for executives and security teams
Plain-English summary
A flaw in TensorFlow can let a local, low-privileged user read memory outside an intended heap buffer by giving illegal inputs to SdcaOptimizerV2. The impact is confidentiality-focused: data exposure is possible, but the sources do not report code execution, data tampering, or service outage.
Executive priority
Prioritize remediation for shared ML platforms or systems processing untrusted model code or tensor inputs. For isolated internal workloads with trusted users only, urgency is lower but upgrading remains appropriate because the defect can expose memory contents.
Technical view
CVE-2021-37672 is a CWE-125 heap out-of-bounds read in tf.raw_ops.SdcaOptimizerV2. The implementation failed to check that example_labels length matched the number of examples. TensorFlow patched this in commit a4e138660270e7599793fa438cd7b2fc2ce215a6, with fixes planned for 2.6.0 and supported backports.
Likely exposure
Exposure is most likely where affected TensorFlow versions run user-controlled ML code, tensors, graphs, notebooks, or training jobs that can invoke SdcaOptimizerV2. Affected ranges include TensorFlow 2.5.0 before 2.5.1, 2.4.x before 2.4.3, and versions before 2.3.4.
Exploitation context
The CVSS vector is local, low complexity, low privileges, and no user interaction. CISA KEV status is false in the provided bundle, and the sources do not state active exploitation. Treat exploitation as plausible in shared ML environments, but not confirmed as observed in the wild.
Researcher notes
The key condition is a mismatch between example_labels length and the number of examples. Analysis should focus on reachable SdcaOptimizerV2 paths and whether an attacker can control arguments locally. Do not assume remote exploitability unless the deployment exposes TensorFlow execution through another service layer.
Mitigation direction
- Upgrade TensorFlow to 2.6.0, or patched branches 2.5.1, 2.4.3, or 2.3.4.
- Pin TensorFlow versions in dependency manifests and rebuild affected containers or runtime images.
- Restrict untrusted users from executing arbitrary TensorFlow raw ops in shared environments.
- Review the TensorFlow advisory and patch commit for vendor-specific upgrade guidance.
Validation and detection
- Inventory TensorFlow versions across applications, notebooks, training workers, and containers.
- Check whether workloads can invoke tf.raw_ops.SdcaOptimizerV2 with user-controlled inputs.
- Verify dependency locks and images use fixed TensorFlow versions.
- Confirm shared ML platforms isolate users and restrict arbitrary local code execution.
Public sources used
Generated from the cited source records. This long-tail analysis has not been individually reviewed by a named human.
Conservative CVE-to-ATT&CK context
These mappings and lookup hints may be relevant to the vulnerability behavior, CWE, affected product, or exposure path. Glexia-inferred context is not an official MITRE, ATT&CK, CWE, or CVE Program mapping.
ATT&CK lookup starting points
Use these exact CWE pages and searches to review the Glexia ATT&CK library from this CVE's weakness and description context.
CWE-125: Exact CWE lookup
Use the exact CWE identifier as the starting point before reviewing related ATT&CK behavior. Open the exact CWE lookup page first, then review the ATT&CK searches from that MITRE weakness context. This is a Glexia lookup hint, not an official ATT&CK mapping.
Open ATT&CK lookupCVE-2021-37672 mapping review
Open the CVE-to-ATT&CK bridge for reviewed, inferred, or future official mappings tied to this CVE.
Open ATT&CK lookup- Severity
- Medium
- CVSS
- 5.5 (3.1)
- Known Exploited
- No
- Published
Vector: CVSS:3.1/AV:L/AC:L/PR:L/UI:N/S:U/C:H/I:N/A:N
CNA and ADP enrichment extracted from CVE v5
These fields come from the CVE record and ADP containers, not from Glexia's Take. They preserve time-varying source decisions such as CISA SSVC, KEV status, CVSS metrics, and provider references.
CVSS vector scores
1 official scoreWe collect every scored CVSS vector available in the official CNA and ADP containers. When more than one version is present, the table keeps the source vectors side by side instead of collapsing them into the highest score.
CVSS:3.1/AV:L/AC:L/PR:L/UI:N/S:U/C:H/I:N/A:N1.83.6Primary CVE scoreVulnerability scoring details
Base CVSS 3.1 score
5.5MediumVector: CVSS:3.1/AV:L/AC:L/PR:L/UI:N/S:U/C:H/I:N/A:N
Source materials
- CVE List V5 sourceCVE List V5
- https://github.com/tensorflow/tensorflow/security/advisories/GHSA-5hj3-vjjf-f5m7CVE reference · x_refsource_CONFIRM
- https://github.com/tensorflow/tensorflow/commit/a4e138660270e7599793fa438cd7b2fc2ce215a6CVE reference · x_refsource_MISC
Products and packages named in the record
CWE details
CWE links open Glexia weakness intelligence pages with official CWE context, developer remediation guidance, and related CVE mappings.
Out-of-bounds Read
Out-of-bounds Read represents a recurring weakness pattern that can create exploitable paths when design, validation, or implementation controls are missing.