LiveActive security incident?Get immediate response
CVE Record

CVE-2021-35958: TensorFlow through 2.5.0 allows attackers to overwrite arbitrary files via a crafted archive when tf.keras....

TensorFlow through 2.5.0 allows attackers to overwrite arbitrary files via a crafted archive when tf.keras.utils.get_file is used with extract=True. NOTE: the vendor's position is that tf.keras.utils.get_file is not intended for untrusted archives

UnknownCVSS not scoredNot KEV-listedUpdated
Discuss exposureBack to CVEs
Glexia's TakeAutomated analysismoderate

Security readout for executives and security teams

Plain-English summary

TensorFlow code that downloads and extracts archives can overwrite files if it processes a crafted archive. Business risk depends on whether your systems pass untrusted archives to tf.keras.utils.get_file with extract=True. The source notes TensorFlow's position that this helper is not intended for untrusted archives.

Executive priority

Treat as a targeted workflow risk, not an internet-wide emergency. Prioritize ML systems that ingest external archives automatically, especially production pipelines or shared environments where file overwrite could affect credentials, models, or application files.

Technical view

CVE-2021-35958 concerns TensorFlow through 2.5.0. When tf.keras.utils.get_file is used with extract=True, a crafted archive may cause arbitrary file overwrite during extraction. The record provides no CVSS score, CWE, confirmed patch, or active-exploitation evidence.

Likely exposure

Exposure is most likely in ML pipelines, notebooks, services, or build jobs using TensorFlow through 2.5.0 and tf.keras.utils.get_file(extract=True) on archives from users, partners, public URLs, or other untrusted sources.

Exploitation context

The provided sources support a crafted-archive arbitrary file overwrite condition, but do not show known exploitation in the wild. KEV status is false. Risk rises where extraction runs with broad filesystem write permissions or in automated pipelines.

Researcher notes

Key uncertainty is remediation status: the bundle names affected TensorFlow versions but does not provide a vendor fix. Focus validation on call-site trust boundaries, extraction privileges, and whether archives can influence paths during extraction. Avoid claiming exploitation without stronger evidence.

Mitigation direction

  • Avoid using extract=True on untrusted archives.
  • Restrict archive sources to trusted, integrity-verified locations.
  • Check TensorFlow or Keras vendor guidance before relying on version upgrades as a fix.
  • Run extraction with least-privilege filesystem permissions.
  • Use a safer extraction workflow that rejects unsafe archive paths.

Validation and detection

  • Inventory TensorFlow versions and flag deployments through 2.5.0.
  • Search code for tf.keras.utils.get_file calls using extract=True.
  • Determine whether archive URLs or files are user-controlled or externally supplied.
  • Review extraction destinations and filesystem write permissions.
  • Document any compensating controls around archive integrity validation.
Prepared
Confidence
medium
Sources
7

Public sources used

Generated from the cited source records. This long-tail analysis has not been individually reviewed by a named human.

Potential ATT&CK relevance

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.

cve · low confidence lookup

CVE-2021-35958 mapping review

Open the CVE-to-ATT&CK bridge for reviewed, inferred, or future official mappings tied to this CVE.

Open ATT&CK lookup
Vulnerability profileCVE Program record
Severity
Unknown
CVSS
Not scored
Known Exploited
No
Published
Official CVE source material

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.

0CVSS vectors
0Timeline events
0ADP providers
6Source links

CVSS and timeline data

No CVSS vectors or timeline events were available in the normalized CVE source material.

Source materials

Affected products

Products and packages named in the record

VendorProductVersion / packageStatus
n/an/an/aListed
Weakness

CWE details

No CWE listed

CWE links open Glexia weakness intelligence pages with official CWE context, developer remediation guidance, and related CVE mappings.