CVE-2025-12058 (GCVE-0-2025-12058)
Vulnerability from cvelistv5
Published
2025-10-29 08:48
Modified
2025-10-29 14:11
Severity ?
5.9 (Medium) - CVSS:4.0/AV:A/AC:H/AT:P/PR:L/UI:P/VC:H/VI:L/VA:L/SC:H/SI:L/SA:L
CWE
  • CWE-502 - Deserialization of Untrusted Data
Summary
The Keras.Model.load_model method, including when executed with the intended security mitigation safe_mode=True, is vulnerable to arbitrary local file loading and Server-Side Request Forgery (SSRF). This vulnerability stems from the way the StringLookup layer is handled during model loading from a specially crafted .keras archive. The constructor for the StringLookup layer accepts a vocabulary argument that can specify a local file path or a remote file path. * Arbitrary Local File Read: An attacker can create a malicious .keras file that embeds a local path in the StringLookup layer's configuration. When the model is loaded, Keras will attempt to read the content of the specified local file and incorporate it into the model state (e.g., retrievable via get_vocabulary()), allowing an attacker to read arbitrary local files on the hosting system. * Server-Side Request Forgery (SSRF): Keras utilizes tf.io.gfile for file operations. Since tf.io.gfile supports remote filesystem handlers (such as GCS and HDFS) and HTTP/HTTPS protocols, the same mechanism can be leveraged to fetch content from arbitrary network endpoints on the server's behalf, resulting in an SSRF condition. The security issue is that the feature allowing external path loading was not properly restricted by the safe_mode=True flag, which was intended to prevent such unintended data access.
References
Impacted products
Vendor Product Version
Keras Keras Version: 0
 Create a notification for this product.
Show details on NVD website

To clipboard 

{
  "containers": {
    "adp": [
      {
        "metrics": [
          {
            "other": {
              "content": {
                "id": "CVE-2025-12058",
                "options": [
                  {
                    "Exploitation": "none"
                  },
                  {
                    "Automatable": "no"
                  },
                  {
                    "Technical Impact": "partial"
                  }
                ],
                "role": "CISA Coordinator",
                "timestamp": "2025-10-29T14:07:04.803189Z",
                "version": "2.0.3"
              },
              "type": "ssvc"
            }
          }
        ],
        "providerMetadata": {
          "dateUpdated": "2025-10-29T14:11:03.027Z",
          "orgId": "134c704f-9b21-4f2e-91b3-4a467353bcc0",
          "shortName": "CISA-ADP"
        },
        "title": "CISA ADP Vulnrichment"
      }
    ],
    "cna": {
      "affected": [
        {
          "defaultStatus": "unaffected",
          "product": "Keras",
          "vendor": "Keras",
          "versions": [
            {
              "lessThan": "3.12.0",
              "status": "affected",
              "version": "0",
              "versionType": "date"
            }
          ]
        }
      ],
      "cpeApplicability": [
        {
          "nodes": [
            {
              "cpeMatch": [
                {
                  "criteria": "cpe:2.3:a:keras:keras:*:*:*:*:*:*:*:*",
                  "versionEndExcluding": "3.12.0",
                  "versionStartIncluding": "0",
                  "vulnerable": true
                }
              ],
              "negate": false,
              "operator": "OR"
            }
          ],
          "operator": "OR"
        }
      ],
      "credits": [
        {
          "lang": "en",
          "type": "finder",
          "value": "Jayashwa Singh Chauhan"
        }
      ],
      "datePublic": "2025-10-17T22:00:00.000Z",
      "descriptions": [
        {
          "lang": "en",
          "supportingMedia": [
            {
              "base64": false,
              "type": "text/html",
              "value": "\u003cp\u003eThe \u003cb\u003e\u003ccode\u003eKeras.Model.load_model\u003c/code\u003e\u003c/b\u003e method, including when executed with the intended security mitigation \u003cb\u003e\u003ccode\u003esafe_mode=True\u003c/code\u003e\u003c/b\u003e, is vulnerable to arbitrary local file loading and Server-Side Request Forgery (SSRF).\u003cbr\u003e\u003c/p\u003e\u003cp\u003eThis vulnerability stems from the way the \u003cb\u003e\u003ccode\u003eStringLookup\u003c/code\u003e\u003c/b\u003e layer is handled during model loading from a specially crafted \u003cb\u003e\u003ccode\u003e.keras\u003c/code\u003e\u003c/b\u003e archive. The constructor for the \u003ccode\u003eStringLookup\u003c/code\u003e layer accepts a \u003ccode\u003evocabulary\u003c/code\u003e argument that can specify a \u003cb\u003elocal file path\u003c/b\u003e or a \u003cb\u003eremote file path\u003c/b\u003e.\u003c/p\u003e\u003col\u003e\u003cli\u003e\u003cp\u003e\u003cb\u003eArbitrary Local File Read:\u003c/b\u003e An attacker can create a malicious \u003ccode\u003e.keras\u003c/code\u003e file that embeds a local path in the \u003ccode\u003eStringLookup\u003c/code\u003e layer\u0027s configuration. When the model is loaded, Keras will attempt to read the content of the specified local file and incorporate it into the model state (e.g., retrievable via \u003ccode\u003eget_vocabulary()\u003c/code\u003e), allowing an attacker to \u003cb\u003eread arbitrary local files\u003c/b\u003e on the hosting system.\u003c/p\u003e\u003c/li\u003e\u003cli\u003e\u003cp\u003e\u003cb\u003eServer-Side Request Forgery (SSRF):\u003c/b\u003e Keras utilizes \u003cb\u003e\u003ccode\u003etf.io.gfile\u003c/code\u003e\u003c/b\u003e for file operations. Since \u003ccode\u003etf.io.gfile\u003c/code\u003e supports remote filesystem handlers (such as GCS and HDFS) and HTTP/HTTPS protocols, the same mechanism can be leveraged to fetch content from \u003cb\u003earbitrary network endpoints\u003c/b\u003e on the server\u0027s behalf, resulting in an SSRF condition.\u003c/p\u003e\u003c/li\u003e\u003c/ol\u003e\u003cp\u003eThe security issue is that the feature allowing external path loading was \u003cb\u003enot properly restricted\u003c/b\u003e by the \u003ccode\u003esafe_mode=True\u003c/code\u003e flag, which was intended to prevent such unintended data access.\u003c/p\u003e\u003cbr\u003e"
            }
          ],
          "value": "The Keras.Model.load_model method, including when executed with the intended security mitigation safe_mode=True, is vulnerable to arbitrary local file loading and Server-Side Request Forgery (SSRF).\n\n\nThis vulnerability stems from the way the StringLookup layer is handled during model loading from a specially crafted .keras archive. The constructor for the StringLookup layer accepts a vocabulary argument that can specify a local file path or a remote file path.\n\n  *  Arbitrary Local File Read: An attacker can create a malicious .keras file that embeds a local path in the StringLookup layer\u0027s configuration. When the model is loaded, Keras will attempt to read the content of the specified local file and incorporate it into the model state (e.g., retrievable via get_vocabulary()), allowing an attacker to read arbitrary local files on the hosting system.\n\n\n  *  Server-Side Request Forgery (SSRF): Keras utilizes tf.io.gfile for file operations. Since tf.io.gfile supports remote filesystem handlers (such as GCS and HDFS) and HTTP/HTTPS protocols, the same mechanism can be leveraged to fetch content from arbitrary network endpoints on the server\u0027s behalf, resulting in an SSRF condition.\n\n\nThe security issue is that the feature allowing external path loading was not properly restricted by the safe_mode=True flag, which was intended to prevent such unintended data access."
        }
      ],
      "impacts": [
        {
          "capecId": "CAPEC-221",
          "descriptions": [
            {
              "lang": "en",
              "value": "CAPEC-221 Data Serialization External Entities Blowup"
            }
          ]
        }
      ],
      "metrics": [
        {
          "cvssV4_0": {
            "Automatable": "NOT_DEFINED",
            "Recovery": "NOT_DEFINED",
            "Safety": "NOT_DEFINED",
            "attackComplexity": "HIGH",
            "attackRequirements": "PRESENT",
            "attackVector": "ADJACENT",
            "baseScore": 5.9,
            "baseSeverity": "MEDIUM",
            "privilegesRequired": "LOW",
            "providerUrgency": "NOT_DEFINED",
            "subAvailabilityImpact": "LOW",
            "subConfidentialityImpact": "HIGH",
            "subIntegrityImpact": "LOW",
            "userInteraction": "PASSIVE",
            "valueDensity": "NOT_DEFINED",
            "vectorString": "CVSS:4.0/AV:A/AC:H/AT:P/PR:L/UI:P/VC:H/VI:L/VA:L/SC:H/SI:L/SA:L",
            "version": "4.0",
            "vulnAvailabilityImpact": "LOW",
            "vulnConfidentialityImpact": "HIGH",
            "vulnIntegrityImpact": "LOW",
            "vulnerabilityResponseEffort": "NOT_DEFINED"
          },
          "format": "CVSS",
          "scenarios": [
            {
              "lang": "en",
              "value": "GENERAL"
            }
          ]
        }
      ],
      "problemTypes": [
        {
          "descriptions": [
            {
              "cweId": "CWE-502",
              "description": "CWE-502 Deserialization of Untrusted Data",
              "lang": "en",
              "type": "CWE"
            }
          ]
        }
      ],
      "providerMetadata": {
        "dateUpdated": "2025-10-29T08:48:29.689Z",
        "orgId": "14ed7db2-1595-443d-9d34-6215bf890778",
        "shortName": "Google"
      },
      "references": [
        {
          "url": "https://github.com/keras-team/keras/security/advisories/GHSA-qg93-c7p6-gg7f"
        },
        {
          "url": "https://github.com/keras-team/keras/pull/21751"
        }
      ],
      "source": {
        "discovery": "UNKNOWN"
      },
      "title": "Vulnerability in Keras Model.load_model Leading to Arbitrary Local File Loading and SSRF",
      "x_generator": {
        "engine": "Vulnogram 0.4.0"
      }
    }
  },
  "cveMetadata": {
    "assignerOrgId": "14ed7db2-1595-443d-9d34-6215bf890778",
    "assignerShortName": "Google",
    "cveId": "CVE-2025-12058",
    "datePublished": "2025-10-29T08:48:29.689Z",
    "dateReserved": "2025-10-22T07:39:21.715Z",
    "dateUpdated": "2025-10-29T14:11:03.027Z",
    "state": "PUBLISHED"
  },
  "dataType": "CVE_RECORD",
  "dataVersion": "5.1",
  "vulnerability-lookup:meta": {
    "nvd": "{\"cve\":{\"id\":\"CVE-2025-12058\",\"sourceIdentifier\":\"cve-coordination@google.com\",\"published\":\"2025-10-29T09:15:35.500\",\"lastModified\":\"2025-10-30T15:03:13.440\",\"vulnStatus\":\"Awaiting Analysis\",\"cveTags\":[],\"descriptions\":[{\"lang\":\"en\",\"value\":\"The Keras.Model.load_model method, including when executed with the intended security mitigation safe_mode=True, is vulnerable to arbitrary local file loading and Server-Side Request Forgery (SSRF).\\n\\n\\nThis vulnerability stems from the way the StringLookup layer is handled during model loading from a specially crafted .keras archive. The constructor for the StringLookup layer accepts a vocabulary argument that can specify a local file path or a remote file path.\\n\\n  *  Arbitrary Local File Read: An attacker can create a malicious .keras file that embeds a local path in the StringLookup layer\u0027s configuration. When the model is loaded, Keras will attempt to read the content of the specified local file and incorporate it into the model state (e.g., retrievable via get_vocabulary()), allowing an attacker to read arbitrary local files on the hosting system.\\n\\n\\n  *  Server-Side Request Forgery (SSRF): Keras utilizes tf.io.gfile for file operations. Since tf.io.gfile supports remote filesystem handlers (such as GCS and HDFS) and HTTP/HTTPS protocols, the same mechanism can be leveraged to fetch content from arbitrary network endpoints on the server\u0027s behalf, resulting in an SSRF condition.\\n\\n\\nThe security issue is that the feature allowing external path loading was not properly restricted by the safe_mode=True flag, which was intended to prevent such unintended data access.\"}],\"metrics\":{\"cvssMetricV40\":[{\"source\":\"cve-coordination@google.com\",\"type\":\"Secondary\",\"cvssData\":{\"version\":\"4.0\",\"vectorString\":\"CVSS:4.0/AV:A/AC:H/AT:P/PR:L/UI:P/VC:H/VI:L/VA:L/SC:H/SI:L/SA:L/E:X/CR:X/IR:X/AR:X/MAV:X/MAC:X/MAT:X/MPR:X/MUI:X/MVC:X/MVI:X/MVA:X/MSC:X/MSI:X/MSA:X/S:X/AU:X/R:X/V:X/RE:X/U:X\",\"baseScore\":5.9,\"baseSeverity\":\"MEDIUM\",\"attackVector\":\"ADJACENT\",\"attackComplexity\":\"HIGH\",\"attackRequirements\":\"PRESENT\",\"privilegesRequired\":\"LOW\",\"userInteraction\":\"PASSIVE\",\"vulnConfidentialityImpact\":\"HIGH\",\"vulnIntegrityImpact\":\"LOW\",\"vulnAvailabilityImpact\":\"LOW\",\"subConfidentialityImpact\":\"HIGH\",\"subIntegrityImpact\":\"LOW\",\"subAvailabilityImpact\":\"LOW\",\"exploitMaturity\":\"NOT_DEFINED\",\"confidentialityRequirement\":\"NOT_DEFINED\",\"integrityRequirement\":\"NOT_DEFINED\",\"availabilityRequirement\":\"NOT_DEFINED\",\"modifiedAttackVector\":\"NOT_DEFINED\",\"modifiedAttackComplexity\":\"NOT_DEFINED\",\"modifiedAttackRequirements\":\"NOT_DEFINED\",\"modifiedPrivilegesRequired\":\"NOT_DEFINED\",\"modifiedUserInteraction\":\"NOT_DEFINED\",\"modifiedVulnConfidentialityImpact\":\"NOT_DEFINED\",\"modifiedVulnIntegrityImpact\":\"NOT_DEFINED\",\"modifiedVulnAvailabilityImpact\":\"NOT_DEFINED\",\"modifiedSubConfidentialityImpact\":\"NOT_DEFINED\",\"modifiedSubIntegrityImpact\":\"NOT_DEFINED\",\"modifiedSubAvailabilityImpact\":\"NOT_DEFINED\",\"Safety\":\"NOT_DEFINED\",\"Automatable\":\"NOT_DEFINED\",\"Recovery\":\"NOT_DEFINED\",\"valueDensity\":\"NOT_DEFINED\",\"vulnerabilityResponseEffort\":\"NOT_DEFINED\",\"providerUrgency\":\"NOT_DEFINED\"}}]},\"weaknesses\":[{\"source\":\"cve-coordination@google.com\",\"type\":\"Secondary\",\"description\":[{\"lang\":\"en\",\"value\":\"CWE-502\"}]}],\"references\":[{\"url\":\"https://github.com/keras-team/keras/pull/21751\",\"source\":\"cve-coordination@google.com\"},{\"url\":\"https://github.com/keras-team/keras/security/advisories/GHSA-qg93-c7p6-gg7f\",\"source\":\"cve-coordination@google.com\"}]}}",
    "vulnrichment": {
      "containers": "{\"adp\": [{\"title\": \"CISA ADP Vulnrichment\", \"metrics\": [{\"other\": {\"type\": \"ssvc\", \"content\": {\"id\": \"CVE-2025-12058\", \"role\": \"CISA Coordinator\", \"options\": [{\"Exploitation\": \"none\"}, {\"Automatable\": \"no\"}, {\"Technical Impact\": \"partial\"}], \"version\": \"2.0.3\", \"timestamp\": \"2025-10-29T14:07:04.803189Z\"}}}], \"providerMetadata\": {\"orgId\": \"134c704f-9b21-4f2e-91b3-4a467353bcc0\", \"shortName\": \"CISA-ADP\", \"dateUpdated\": \"2025-10-29T14:10:59.454Z\"}}], \"cna\": {\"title\": \"Vulnerability in Keras Model.load_model Leading to Arbitrary Local File Loading and SSRF\", \"source\": {\"discovery\": \"UNKNOWN\"}, \"credits\": [{\"lang\": \"en\", \"type\": \"finder\", \"value\": \"Jayashwa Singh Chauhan\"}], \"impacts\": [{\"capecId\": \"CAPEC-221\", \"descriptions\": [{\"lang\": \"en\", \"value\": \"CAPEC-221 Data Serialization External Entities Blowup\"}]}], \"metrics\": [{\"format\": \"CVSS\", \"cvssV4_0\": {\"Safety\": \"NOT_DEFINED\", \"version\": \"4.0\", \"Recovery\": \"NOT_DEFINED\", \"baseScore\": 5.9, \"Automatable\": \"NOT_DEFINED\", \"attackVector\": \"ADJACENT\", \"baseSeverity\": \"MEDIUM\", \"valueDensity\": \"NOT_DEFINED\", \"vectorString\": \"CVSS:4.0/AV:A/AC:H/AT:P/PR:L/UI:P/VC:H/VI:L/VA:L/SC:H/SI:L/SA:L\", \"providerUrgency\": \"NOT_DEFINED\", \"userInteraction\": \"PASSIVE\", \"attackComplexity\": \"HIGH\", \"attackRequirements\": \"PRESENT\", \"privilegesRequired\": \"LOW\", \"subIntegrityImpact\": \"LOW\", \"vulnIntegrityImpact\": \"LOW\", \"subAvailabilityImpact\": \"LOW\", \"vulnAvailabilityImpact\": \"LOW\", \"subConfidentialityImpact\": \"HIGH\", \"vulnConfidentialityImpact\": \"HIGH\", \"vulnerabilityResponseEffort\": \"NOT_DEFINED\"}, \"scenarios\": [{\"lang\": \"en\", \"value\": \"GENERAL\"}]}], \"affected\": [{\"vendor\": \"Keras\", \"product\": \"Keras\", \"versions\": [{\"status\": \"affected\", \"version\": \"0\", \"lessThan\": \"3.12.0\", \"versionType\": \"date\"}], \"defaultStatus\": \"unaffected\"}], \"datePublic\": \"2025-10-17T22:00:00.000Z\", \"references\": [{\"url\": \"https://github.com/keras-team/keras/security/advisories/GHSA-qg93-c7p6-gg7f\"}, {\"url\": \"https://github.com/keras-team/keras/pull/21751\"}], \"x_generator\": {\"engine\": \"Vulnogram 0.4.0\"}, \"descriptions\": [{\"lang\": \"en\", \"value\": \"The Keras.Model.load_model method, including when executed with the intended security mitigation safe_mode=True, is vulnerable to arbitrary local file loading and Server-Side Request Forgery (SSRF).\\n\\n\\nThis vulnerability stems from the way the StringLookup layer is handled during model loading from a specially crafted .keras archive. The constructor for the StringLookup layer accepts a vocabulary argument that can specify a local file path or a remote file path.\\n\\n  *  Arbitrary Local File Read: An attacker can create a malicious .keras file that embeds a local path in the StringLookup layer\u0027s configuration. When the model is loaded, Keras will attempt to read the content of the specified local file and incorporate it into the model state (e.g., retrievable via get_vocabulary()), allowing an attacker to read arbitrary local files on the hosting system.\\n\\n\\n  *  Server-Side Request Forgery (SSRF): Keras utilizes tf.io.gfile for file operations. Since tf.io.gfile supports remote filesystem handlers (such as GCS and HDFS) and HTTP/HTTPS protocols, the same mechanism can be leveraged to fetch content from arbitrary network endpoints on the server\u0027s behalf, resulting in an SSRF condition.\\n\\n\\nThe security issue is that the feature allowing external path loading was not properly restricted by the safe_mode=True flag, which was intended to prevent such unintended data access.\", \"supportingMedia\": [{\"type\": \"text/html\", \"value\": \"\u003cp\u003eThe \u003cb\u003e\u003ccode\u003eKeras.Model.load_model\u003c/code\u003e\u003c/b\u003e method, including when executed with the intended security mitigation \u003cb\u003e\u003ccode\u003esafe_mode=True\u003c/code\u003e\u003c/b\u003e, is vulnerable to arbitrary local file loading and Server-Side Request Forgery (SSRF).\u003cbr\u003e\u003c/p\u003e\u003cp\u003eThis vulnerability stems from the way the \u003cb\u003e\u003ccode\u003eStringLookup\u003c/code\u003e\u003c/b\u003e layer is handled during model loading from a specially crafted \u003cb\u003e\u003ccode\u003e.keras\u003c/code\u003e\u003c/b\u003e archive. The constructor for the \u003ccode\u003eStringLookup\u003c/code\u003e layer accepts a \u003ccode\u003evocabulary\u003c/code\u003e argument that can specify a \u003cb\u003elocal file path\u003c/b\u003e or a \u003cb\u003eremote file path\u003c/b\u003e.\u003c/p\u003e\u003col\u003e\u003cli\u003e\u003cp\u003e\u003cb\u003eArbitrary Local File Read:\u003c/b\u003e An attacker can create a malicious \u003ccode\u003e.keras\u003c/code\u003e file that embeds a local path in the \u003ccode\u003eStringLookup\u003c/code\u003e layer\u0027s configuration. When the model is loaded, Keras will attempt to read the content of the specified local file and incorporate it into the model state (e.g., retrievable via \u003ccode\u003eget_vocabulary()\u003c/code\u003e), allowing an attacker to \u003cb\u003eread arbitrary local files\u003c/b\u003e on the hosting system.\u003c/p\u003e\u003c/li\u003e\u003cli\u003e\u003cp\u003e\u003cb\u003eServer-Side Request Forgery (SSRF):\u003c/b\u003e Keras utilizes \u003cb\u003e\u003ccode\u003etf.io.gfile\u003c/code\u003e\u003c/b\u003e for file operations. Since \u003ccode\u003etf.io.gfile\u003c/code\u003e supports remote filesystem handlers (such as GCS and HDFS) and HTTP/HTTPS protocols, the same mechanism can be leveraged to fetch content from \u003cb\u003earbitrary network endpoints\u003c/b\u003e on the server\u0027s behalf, resulting in an SSRF condition.\u003c/p\u003e\u003c/li\u003e\u003c/ol\u003e\u003cp\u003eThe security issue is that the feature allowing external path loading was \u003cb\u003enot properly restricted\u003c/b\u003e by the \u003ccode\u003esafe_mode=True\u003c/code\u003e flag, which was intended to prevent such unintended data access.\u003c/p\u003e\u003cbr\u003e\", \"base64\": false}]}], \"problemTypes\": [{\"descriptions\": [{\"lang\": \"en\", \"type\": \"CWE\", \"cweId\": \"CWE-502\", \"description\": \"CWE-502 Deserialization of Untrusted Data\"}]}], \"cpeApplicability\": [{\"nodes\": [{\"negate\": false, \"cpeMatch\": [{\"criteria\": \"cpe:2.3:a:keras:keras:*:*:*:*:*:*:*:*\", \"vulnerable\": true, \"versionEndExcluding\": \"3.12.0\", \"versionStartIncluding\": \"0\"}], \"operator\": \"OR\"}], \"operator\": \"OR\"}], \"providerMetadata\": {\"orgId\": \"14ed7db2-1595-443d-9d34-6215bf890778\", \"shortName\": \"Google\", \"dateUpdated\": \"2025-10-29T08:48:29.689Z\"}}}",
      "cveMetadata": "{\"cveId\": \"CVE-2025-12058\", \"state\": \"PUBLISHED\", \"dateUpdated\": \"2025-10-29T14:11:03.027Z\", \"dateReserved\": \"2025-10-22T07:39:21.715Z\", \"assignerOrgId\": \"14ed7db2-1595-443d-9d34-6215bf890778\", \"datePublished\": \"2025-10-29T08:48:29.689Z\", \"assignerShortName\": \"Google\"}",
      "dataType": "CVE_RECORD",
      "dataVersion": "5.1"
    }
  }
}


Log in or create an account to share your comment.




Tags
Taxonomy of the tags.




Sightings

Author Source Type Date

Nomenclature

  • Seen: The vulnerability was mentioned, discussed, or seen somewhere by the user.
  • Confirmed: The vulnerability is confirmed from an analyst perspective.
  • Published Proof of Concept: A public proof of concept is available for this vulnerability.
  • Exploited: This vulnerability was exploited and seen by the user reporting the sighting.
  • Patched: This vulnerability was successfully patched by the user reporting the sighting.
  • Not exploited: This vulnerability was not exploited or seen by the user reporting the sighting.
  • Not confirmed: The user expresses doubt about the veracity of the vulnerability.
  • Not patched: This vulnerability was not successfully patched by the user reporting the sighting.