CVE-2025-9905

HIGHNVD 7.37.3
EchelonGraph scoreMEDIUM confidence

“—” beside the score: it is equal to NVD's CVSS score.

This high-severity CVE scores 7.3 under NVD CVSS v3. EPSS exploit probability: 0.2% (12th percentile of EPSS-scored CVEs). No GitHub Security Advisory CVSS is on record for this CVE.

Triggered by: Headline CVSS from NVD CVSS v3
Sources: EPSS, NVD CVSS v3
Trending — 5 sources updated this week

Also known asEUVD-2025-30277 (ENISA EUVD) · GHSA-36rr-ww3j-vrjv (GitHub Security Advisories) · PYSEC-2025-123 (OSV.dev)Source: ENISA EUVD

7.3EG
EchelonGraph verdictPlan a fixRemediate within your standard update timelines (the SSVC decision on this page).
  • High severity, but no confirmed exploitation yet
CISA-KEV: Not listedEPSS PROB: 0.2%CVSS: 7.3Exploit: None knownExposed services: Not assessed

A fix is available — apply it.

The Keras Model.load_model method can be exploited to achieve arbitrary code execution, even with safe_mode=True.

One can create a specially crafted .h5/.hdf5 model archive that, when loaded via Model.load_model, will trigger arbitrary code to be executed.

This is achieved by crafting a special .h5 archive file that uses the Lambda layer feature of keras which allows arbitrary Python code in the form of pickled code. The vulnerability comes from the fact that the safe_mode=True option is not honored when reading .h5 archives.

Note that the .h5/.hdf5 format is a legacy format supported by Keras 3 for backwards compatibility.

CVSS v3
7.3
EG Score
7.3HIGHmedium confidence
EG Risk
33
EG Risk 33/100

EG Risk is EchelonGraph's 0–100 priority score. It fuses impact severity (EG Severity, the CVSS base score, never the KEV-floored or EPSS-lifted EG Score) with real-world exploitation and automatability, under weights EchelonGraph set by hand, so you can rank equal-severity CVEs and fix the most dangerous first. Higher = act sooner.

How it’s computed (value × EchelonGraph’s weight)
Severity73% × 45%
Impact severity (EG Severity): CVSS 7.3 of 10. Never the KEV-floored or EPSS-lifted EG Score.
Exploitation0.2% × 40%
FIRST EPSS: the 30-day exploitation probability when EchelonGraph scored this CVE.
Automatability0% × 15%
CISA's SSVC Automatable value: no (Vulnrichment).
SSVC (CISA's decision table, applied by EchelonGraph): Track at low or medium mission impact; Track* at high (mission-essential systems).
Exploitation none (CISA Vulnrichment) · Automatable no (CISA Vulnrichment) · Technical impact total (CISA Vulnrichment). All three are CISA's SSVC values (Vulnrichment). EchelonGraph applied CISA's decision table to them; CISA publishes SSVC inputs and the table, not a decision for each CVE.
Action: A fix is available. Apply it within your standard update timelines.
EPSS PROB
0.2%
EPSS %ILE
12th
KEV
Not listed

What the EG score is made of (scorer v4.0: the three questions the 0–10 score mixes, published apart · methodology)

The EG score (7.3): severity, escalated by confirmed exploitation. This CVE's 7.3 is its impact severity (EG Severity), not escalated.

EG Severity
7.3 HIGHCVSS-B, impact only: no KEV listing, EPSS or EG-KEV tier raises it.NVD CVSS v3 · CVSS v3.0 or v3.1
EG Likelihood
0.2% within 30 daysSource: FIRST EPSS (model v2026.06.15, score date 2026-10-31). FIRST's exploitation probability, not an EchelonGraph model.
Confirmed exploitation
None on record (CISA KEV, EG-KEV Tier 1)Reported apart from the likelihood: an observed fact, not a probability.
EG Priority
SSVC Track at medium mission impactTrack at low or medium mission impact; Track* at high (mission-essential systems).CISA BOD 26-04: 14 days if the affected system is publicly exposed; Fix on system upgrade if it is not.Each input and its source, and whom BOD 26-04 binds: the SSVC and BOD 26-04 blocks below.

SSVC (CISA's decision table, applied by EchelonGraph)Track at low or medium mission impact; Track* at high (mission-essential systems).

A fix is available. Apply it within your standard update timelines.

Exploitation none (CISA Vulnrichment) · Automatable no (CISA Vulnrichment) · Technical impact total (CISA Vulnrichment). All three are CISA's SSVC values (Vulnrichment). EchelonGraph applied CISA's decision table to them; CISA publishes SSVC inputs and the table, not a decision for each CVE. Mission impact is CISA's Mission & Well-being decision point, and only you can judge it. CISA's table makes it high when Mission Prevalence is Essential — the vulnerable component "directly provides capabilities that constitute at least one MEF for at least one entity" (MEF: mission essential function) — or when Public Well-Being Impact is Irreversible: "multiple fatalities are likely", the cyber-physical system "is likely lost or destroyed", "extreme or serious externalities" are imposed on other parties, or social systems such as elections or the financial grid "are destabilized and potentially collapse". CISA's decision table

BOD 26-04 (CISA's remediation timeline, Table 1)14 days if the affected system is publicly exposed; Fix on system upgrade if it is not.

The score's inputs: In CISA KEV no (CISA's KEV catalog) · Automatable no (CISA Vulnrichment) · Technical impact total (CISA Vulnrichment). Whether your system is publicly exposed is yours to answer. BOD 26-04 binds US Federal Civilian Executive Branch agencies; for anyone else it is CISA's published timeline, a reference rather than an obligation. The directive · how EchelonGraph applies it

Published

September 19, 2025

Last Modified

October 19, 2026

Advisory Details (2)

Auto-updated Oct 17, 2026
🔬 Proof of concept available. Patch available. Sources: github_pr, github.
GitHub advisoryMentions a patch (unverified)🟡 PoC Available

The Keras `Model.load_model` method **silently** ignores `safe_mode=True` and allows arbitrary code execution when a `.h5`/`.hdf5` file is loaded. · Advisory · keras-team/keras · GitHub

https://github.com/keras-team/keras/security/advisories/GHSA-36rr-ww3j-vrjv
GitHub pull request Patch Available

Propagate `safe_mode` flag to legacy h5 loading code.

Patch available: keras-team/keras v3.12.0 (PR #21602 merged 2025-08-20)

https://github.com/keras-team/keras/pull/21602

Patch Availability(3)

Vendor / EcosystemFix (by version range) / PatchReleasedSource
redhatrhoai/odh-modelmesh-runtime-adapter-rhel9:v2.25.1-17648250962025-12-17redhat
redhatrhoai/odh-modelmesh-runtime-adapter-rhel9:v2.22.3-17645945082025-12-04redhat
pipkeras @ 3.11.3 (fixes >= 3.0.0, < 3.11.3)—ghsa

Patches are aggregated from vendor advisories (Red Hat, Microsoft, Cisco, GitHub) and package ecosystems (OSV, GHSA). Multiple rows for the same upstream release have been deduplicated.

Affected Packages

(2 across 2 ecosystems)
Debian:11(1)
PackageVulnerable rangeFix by version rangeDependents
keras2.3.1+dfsg-3, 2.3.1+dfsg2-1
  • every version on: no fix on record
—
PyPI(1)
PackageVulnerable rangeFix by version rangeDependents
keras3.0.0 ... 3.9.2 (27 versions)
  • 3.0.0 up to 3.11.3: fixed in 3.11.3
—

Weakness Classification(1)

MITRE Common Weakness Enumeration — the root-cause categories this CVE belongs to.

Data Freshness Timeline

(refreshed 12× in last 7d / 47× in last 30d)

Each row is a time one of our pipelines fetched, updated or re-scored this CVE. For example, "NVD update" means NVD published or revised its analysis for this CVE; "MITRE cvelistV5" means we ingested or refreshed it from the CNA feed. An arrow marks a change the record states ("CVSS v3 → 9.1"); EPSS and GHSA rows show the values recorded at that time, not what changed (EPSS values are rounded; hover one for the exact figures). Most recent first.

Showing the most recent 100 of 181 refreshes on record for this CVE.

A repeat, a row identical to one the same pipeline wrote less than 10 seconds earlier, is not a separate refresh and is not counted. None are among the latest 100 rows. The 81 oldest rows on record are not shown and were not checked, so the counts can still include repeats.

  1. 2026-11-02 16:43 UTCEPSS rescoreEPSS 0.2% · 12th percentile
  2. 2026-11-01 16:33 UTCEPSS rescoreEPSS 0.2% · 12th percentile
  3. 2026-11-01 08:01 UTCGHSA enrichmentGHSA-36rr-ww3j-vrjv · GitHub severity HIGH
  4. 2026-11-01 08:01 UTCEG score recompute
  5. 2026-10-31 16:31 UTCEPSS rescoreEPSS 0.2% · 12th percentile
  6. 2026-10-30 16:33 UTCEPSS rescoreEPSS 0.2% · 12th percentile
  7. 2026-10-30 00:44 UTCEG score recompute
  8. 2026-10-30 00:44 UTCVendor advisory
  9. 2026-10-29 20:55 UTCOSV refresh
  10. 2026-10-29 16:32 UTCEPSS rescoreEPSS 0.2% · 12th percentile
  11. 2026-10-28 16:31 UTCEPSS rescoreEPSS 0.2% · 12th percentile
  12. 2026-10-27 16:31 UTCEPSS rescoreEPSS 0.2% · 12th percentile
  13. 2026-10-26 16:30 UTCEPSS rescoreEPSS 0.2% · 12th percentile
  14. 2026-10-25 16:32 UTCEPSS rescoreEPSS 0.2% · 12th percentile
  15. 2026-10-25 15:59 UTCEG score recompute
  16. 2026-10-24 16:31 UTCEPSS rescoreEPSS 0.2% · 12th percentile
  17. 2026-10-23 16:31 UTCEPSS rescoreEPSS 0.2% · 12th percentile
  18. 2026-10-22 16:33 UTCEPSS rescoreEPSS 0.2% · 12th percentile
  19. 2026-10-22 11:31 UTCEG score recompute
  20. 2026-10-22 11:31 UTCVendor advisory
  21. 2026-10-21 16:31 UTCEPSS rescoreEPSS 0.2% · 12th percentile
  22. 2026-10-20 16:31 UTCEPSS rescoreEPSS 0.2% · 12th percentile
  23. 2026-10-19 16:31 UTCEPSS rescoreEPSS 0.2% · 12th percentile
  24. 2026-10-19 16:29 UTCEG score recompute
  25. 2026-10-19 16:29 UTCVendor advisory
Show 75 more
  1. 2026-10-18 16:31 UTCEPSS rescoreEPSS 0.2% · 11th percentile
  2. 2026-10-18 05:13 UTCEG score recompute
  3. 2026-10-17 20:50 UTCEG score recompute
  4. 2026-10-17 20:50 UTCVendor advisory
  5. 2026-10-17 16:33 UTCEPSS rescoreEPSS 0.2% · 11th percentile
  6. 2026-10-16 16:40 UTCEPSS rescoreEPSS 0.2% · 11th percentile
  7. 2026-10-15 16:31 UTCEPSS rescoreEPSS 0.2% · 11th percentile
  8. 2026-10-14 16:31 UTCEPSS rescoreEPSS 0.2% · 11th percentile
  9. 2026-10-13 16:33 UTCEPSS rescoreEPSS 0.2% · 11th percentile
  10. 2026-10-12 16:43 UTCEG score recompute
  11. 2026-10-12 16:33 UTCEPSS rescoreEPSS 0.2% · 11th percentile
  12. 2026-10-11 20:18 UTCEG score recompute
  13. 2026-10-11 18:30 UTCEG score recompute
  14. 2026-10-11 15:04 UTCEG score recompute
  15. 2026-10-11 13:47 UTCEPSS rescoreEPSS 0.2% · 11th percentile
  16. 2026-10-10 19:59 UTCEPSS rescoreEPSS 0.2% · 11th percentile
  17. 2026-10-09 20:06 UTCEPSS rescoreEPSS 0.2% · 11th percentile
  18. 2026-10-08 17:11 UTCEPSS rescoreEPSS 0.2% · 11th percentile
  19. 2026-10-07 23:47 UTCEPSS rescoreEPSS 0.2% · 11th percentile
  20. 2026-10-06 15:17 UTCEPSS rescoreEPSS 0.2% · 11th percentile
  21. 2026-10-05 17:42 UTCEPSS rescoreEPSS 0.2% · 11th percentile
  22. 2026-10-04 23:21 UTCEPSS rescoreEPSS 0.2% · 11th percentile
  23. 2026-10-03 14:25 UTCEPSS rescoreEPSS 0.2% · 11th percentile
  24. 2026-10-02 04:40 UTCOSV refresh
  25. 2026-10-01 19:49 UTCEPSS rescoreEPSS 0.2% · 11th percentile
  26. 2026-09-30 15:02 UTCEPSS rescoreEPSS 0.2% · 11th percentile
  27. 2026-09-29 15:20 UTCEPSS rescoreEPSS 0.2% · 10th percentile
  28. 2026-09-28 13:50 UTCEPSS rescoreEPSS 0.2% · 10th percentile
  29. 2026-09-26 15:58 UTCEPSS rescoreEPSS 0.2% · 10th percentile
  30. 2026-09-23 13:47 UTCEPSS rescoreEPSS 0.2% · 12th percentile
  31. 2026-09-20 20:15 UTCEPSS rescoreEPSS 0.2% · 12th percentile
  32. 2026-09-18 19:27 UTCEPSS rescoreEPSS 0.2% · 12th percentile
  33. 2026-09-16 05:14 UTCEPSS rescoreEPSS 0.2% · 12th percentile
  34. 2026-09-13 16:46 UTCEPSS rescoreEPSS 0.2% · 12th percentile
  35. 2026-09-12 23:06 UTCOSV refresh
  36. 2026-09-12 15:00 UTCEPSS rescoreEPSS 0.2% · 12th percentile
  37. 2026-09-11 14:52 UTCEPSS rescoreEPSS 0.2% · 12th percentile
  38. 2026-09-10 09:34 UTCEPSS rescoreEPSS 0.2% · 12th percentile
  39. 2026-09-08 22:00 UTCEPSS rescoreEPSS 0.2% · 12th percentile
  40. 2026-09-07 16:01 UTCEPSS rescoreEPSS 0.2% · 12th percentile
  41. 2026-09-06 13:47 UTCEPSS rescoreEPSS 0.2% · 12th percentile
  42. 2026-09-04 05:06 UTCEPSS rescoreEPSS 0.2% · 12th percentile
  43. 2026-09-02 14:11 UTCEPSS rescoreEPSS 0.2% · 12th percentile
  44. 2026-09-01 13:53 UTCEPSS rescoreEPSS 0.2% · 12th percentile
  45. 2026-09-01 04:39 UTCEPSS rescoreEPSS 0.2% · 12th percentile
  46. 2026-08-30 19:17 UTCEPSS rescoreEPSS 0.2% · 12th percentile
  47. 2026-08-30 01:21 UTCEPSS rescoreEPSS 0.2% · 12th percentile
  48. 2026-08-28 21:40 UTCEPSS rescoreEPSS 0.2% · 12th percentile
  49. 2026-08-27 14:24 UTCEPSS rescoreEPSS 0.2% · 12th percentile
  50. 2026-08-25 13:48 UTCEPSS rescoreEPSS 0.2% · 12th percentile
  51. 2026-08-25 03:26 UTCOSV refresh
  52. 2026-08-23 00:19 UTCEPSS rescoreEPSS 0.2% · 12th percentile
  53. 2026-08-21 23:48 UTCEPSS rescoreEPSS 0.2% · 12th percentile
  54. 2026-08-20 22:55 UTCEPSS rescoreEPSS 0.2% · 12th percentile
  55. 2026-08-19 17:03 UTCEPSS rescoreEPSS 0.2% · 10th percentile
  56. 2026-08-18 13:48 UTCEPSS rescoreEPSS 0.2% · 10th percentile
  57. 2026-08-16 14:56 UTCEPSS rescoreEPSS 0.2% · 10th percentile
  58. 2026-08-16 02:14 UTCEPSS rescoreEPSS 0.2% · 10th percentile
  59. 2026-08-15 01:30 UTCEPSS rescoreEPSS 0.2% · 10th percentile
  60. 2026-08-13 22:00 UTCEPSS rescoreEPSS 0.2% · 10th percentile
  61. 2026-08-12 13:50 UTCEPSS rescoreEPSS 0.2% · 10th percentile
  62. 2026-08-11 13:42 UTCEPSS rescoreEPSS 0.2% · 10th percentile
  63. 2026-08-10 23:59 UTCEPSS rescoreEPSS 0.2% · 10th percentile
  64. 2026-08-09 13:46 UTCEPSS rescoreEPSS 0.2% · 10th percentile
  65. 2026-08-08 20:49 UTCOSV refresh
  66. 2026-08-08 16:37 UTCEPSS rescoreEPSS 0.2% · 10th percentile
  67. 2026-08-06 13:46 UTCEPSS rescoreEPSS 0.2% · 10th percentile
  68. 2026-08-05 19:17 UTCEPSS rescoreEPSS 0.2% · 10th percentile
  69. 2026-08-04 15:09 UTCEPSS rescoreEPSS 0.2% · 10th percentile
  70. 2026-08-04 10:38 UTCEPSS rescoreEPSS 0.2% · 10th percentile
  71. 2026-08-03 10:35 UTCEPSS rescoreEPSS 0.2% · 10th percentile
  72. 2026-08-02 02:26 UTCEPSS rescoreEPSS 0.2% · 10th percentile
  73. 2026-08-01 04:15 UTCEPSS rescoreEPSS 0.2% · 10th percentile
  74. 2026-07-30 16:27 UTCEPSS rescoreEPSS 0.2% · 10th percentile
  75. 2026-07-30 01:30 UTCEPSS rescoreEPSS 0.2% · 10th percentile

Frequently asked(5)

What is CVE-2025-9905?
CVE-2025-9905 is a high-severity vulnerability; its CVE record was published on September 19, 2025. The Keras Model.load_model method can be exploited to achieve arbitrary code execution, even with safe_mode=True. One can create a specially crafted .h5/.hdf5 model archive that, when loaded via Model.load_model, will trigger arbitrary code to be executed. This is achieved by crafting a special .h5…
When was the CVE record for CVE-2025-9905 published?
The CVE record for CVE-2025-9905 was published on September 19, 2025. That is the record's publication date; the vulnerability itself may have been made public earlier. EchelonGraph's copy of the record carries a last-modified date of October 19, 2026; the record at its source may have been updated since.
Is CVE-2025-9905 actively exploited?
CVE-2025-9905 is not currently on CISA's Known Exploited Vulnerabilities catalog. FIRST EPSS estimates a 0.2% probability of exploitation in the next 30 days (12th percentile of EPSS-scored CVEs).
What is the CVSS score of CVE-2025-9905?
CVE-2025-9905 has a CVSS v3 base score of 7.3 (NVD).
How do I remediate CVE-2025-9905?
A fix for CVE-2025-9905 is available: update to the fixed version the vendor names in its advisory. The vendor advisories EchelonGraph has for CVE-2025-9905 are linked in the Vendor Advisories panel on this page.

Dependency Blast Radius

See which npm, PyPI, Go, and Maven packages are affected by CVE-2025-9905

Explore →

Is Your Infrastructure Affected by CVE-2025-9905?

EchelonGraph automatically scans your cloud infrastructure and maps CVE exposure using blast radius analysis.