{"api_version":"1","generated_at":"2026-07-23T16:47:57+00:00","cve":"CVE-2021-29521","urls":{"html":"https://cve.report/CVE-2021-29521","api":"https://cve.report/api/cve/CVE-2021-29521.json","docs":"https://cve.report/api","cve_org":"https://www.cve.org/CVERecord?id=CVE-2021-29521","nvd":"https://nvd.nist.gov/vuln/detail/CVE-2021-29521"},"summary":{"title":"CVE-2021-29521","description":"TensorFlow is an end-to-end open source platform for machine learning. Specifying a negative dense shape in `tf.raw_ops.SparseCountSparseOutput` results in a segmentation fault being thrown out from the standard library as `std::vector` invariants are broken. This is because the implementation(https://github.com/tensorflow/tensorflow/blob/8f7b60ee8c0206a2c99802e3a4d1bb55d2bc0624/tensorflow/core/kernels/count_ops.cc#L199-L213) assumes the first element of the dense shape is always positive and uses it to initialize a `BatchedMap<T>` (i.e., `std::vector<absl::flat_hash_map<int64,T>>`(https://github.com/tensorflow/tensorflow/blob/8f7b60ee8c0206a2c99802e3a4d1bb55d2bc0624/tensorflow/core/kernels/count_ops.cc#L27)) data structure. If the `shape` tensor has more than one element, `num_batches` is the first value in `shape`. Ensuring that the `dense_shape` argument is a valid tensor shape (that is, all elements are non-negative) solves this issue. The fix will be included in TensorFlow 2.5.0. We will also cherrypick this commit on TensorFlow 2.4.2 and TensorFlow 2.3.3.","state":"PUBLIC","assigner":"security-advisories@github.com","published_at":"2021-05-14 20:15:00","updated_at":"2021-05-20 17:19:00"},"problem_types":["CWE-131"],"metrics":[],"references":[{"url":"https://github.com/tensorflow/tensorflow/commit/c57c0b9f3a4f8684f3489dd9a9ec627ad8b599f5","name":"https://github.com/tensorflow/tensorflow/commit/c57c0b9f3a4f8684f3489dd9a9ec627ad8b599f5","refsource":"MISC","tags":[],"title":"Fix the segfault in `tf.raw_ops.SparseCountSparseOutput`. · tensorflow/tensorflow@c57c0b9 · GitHub","mime":"text/html","httpstatus":"200","archivestatus":"404"},{"url":"https://github.com/tensorflow/tensorflow/security/advisories/GHSA-hr84-fqvp-48mm","name":"https://github.com/tensorflow/tensorflow/security/advisories/GHSA-hr84-fqvp-48mm","refsource":"CONFIRM","tags":[],"title":"Segfault in `SparseCountSparseOutput` · Advisory · tensorflow/tensorflow · GitHub","mime":"text/html","httpstatus":"200","archivestatus":"404"},{"url":"https://www.cve.org/CVERecord?id=CVE-2021-29521","name":"CVE Program record","refsource":"CVE.ORG","tags":["canonical"]},{"url":"https://nvd.nist.gov/vuln/detail/CVE-2021-29521","name":"NVD vulnerability detail","refsource":"NVD","tags":["canonical","analysis"]}],"affected":[],"timeline":[],"solutions":[],"workarounds":[],"exploits":[],"credits":[],"nvd_cpes":[{"cve_year":"2021","cve_id":"29521","vulnerable":"1","versionEndIncluding":"","cpe1":"cpe","cpe2":"2.3","cpe3":"a","cpe4":"google","cpe5":"tensorflow","cpe6":"*","cpe7":"*","cpe8":"*","cpe9":"*","cpe10":"*","cpe11":"*","cpe12":"*","cpe13":"*"}],"vendor_comments":[],"enrichments":{"kev":null,"epss":null,"legacy_qids":[{"cve":"CVE-2021-29521","qid":"982525","title":"Python (pip) Security Update for tensorflow-gpu (GHSA-hr84-fqvp-48mm)"}]},"source_records":{"cve_program":{"CVE_data_meta":{"ASSIGNER":"security-advisories@github.com","ID":"CVE-2021-29521","STATE":"PUBLIC","TITLE":"Segfault in SparseCountSparseOutput"},"affects":{"vendor":{"vendor_data":[{"product":{"product_data":[{"product_name":"tensorflow","version":{"version_data":[{"version_value":"< 2.3.3"},{"version_value":">= 2.4.0, < 2.4.2"}]}}]},"vendor_name":"tensorflow"}]}},"data_format":"MITRE","data_type":"CVE","data_version":"4.0","description":{"description_data":[{"lang":"eng","value":"TensorFlow is an end-to-end open source platform for machine learning. Specifying a negative dense shape in `tf.raw_ops.SparseCountSparseOutput` results in a segmentation fault being thrown out from the standard library as `std::vector` invariants are broken. This is because the implementation(https://github.com/tensorflow/tensorflow/blob/8f7b60ee8c0206a2c99802e3a4d1bb55d2bc0624/tensorflow/core/kernels/count_ops.cc#L199-L213) assumes the first element of the dense shape is always positive and uses it to initialize a `BatchedMap<T>` (i.e., `std::vector<absl::flat_hash_map<int64,T>>`(https://github.com/tensorflow/tensorflow/blob/8f7b60ee8c0206a2c99802e3a4d1bb55d2bc0624/tensorflow/core/kernels/count_ops.cc#L27)) data structure. If the `shape` tensor has more than one element, `num_batches` is the first value in `shape`. Ensuring that the `dense_shape` argument is a valid tensor shape (that is, all elements are non-negative) solves this issue. The fix will be included in TensorFlow 2.5.0. We will also cherrypick this commit on TensorFlow 2.4.2 and TensorFlow 2.3.3."}]},"impact":{"cvss":{"attackComplexity":"HIGH","attackVector":"LOCAL","availabilityImpact":"LOW","baseScore":2.5,"baseSeverity":"LOW","confidentialityImpact":"NONE","integrityImpact":"NONE","privilegesRequired":"LOW","scope":"UNCHANGED","userInteraction":"NONE","vectorString":"CVSS:3.1/AV:L/AC:H/PR:L/UI:N/S:U/C:N/I:N/A:L","version":"3.1"}},"problemtype":{"problemtype_data":[{"description":[{"lang":"eng","value":"CWE-131: Incorrect Calculation of Buffer Size"}]}]},"references":{"reference_data":[{"name":"https://github.com/tensorflow/tensorflow/security/advisories/GHSA-hr84-fqvp-48mm","refsource":"CONFIRM","url":"https://github.com/tensorflow/tensorflow/security/advisories/GHSA-hr84-fqvp-48mm"},{"name":"https://github.com/tensorflow/tensorflow/commit/c57c0b9f3a4f8684f3489dd9a9ec627ad8b599f5","refsource":"MISC","url":"https://github.com/tensorflow/tensorflow/commit/c57c0b9f3a4f8684f3489dd9a9ec627ad8b599f5"}]},"source":{"advisory":"GHSA-hr84-fqvp-48mm","discovery":"UNKNOWN"}},"nvd":{"publishedDate":"2021-05-14 20:15:00","lastModifiedDate":"2021-05-20 17:19:00","problem_types":["CWE-131"],"metrics":{"baseMetricV3":{"cvssV3":{"version":"3.1","vectorString":"CVSS:3.1/AV:L/AC:L/PR:L/UI:N/S:U/C:N/I:N/A:H","attackVector":"LOCAL","attackComplexity":"LOW","privilegesRequired":"LOW","userInteraction":"NONE","scope":"UNCHANGED","confidentialityImpact":"NONE","integrityImpact":"NONE","availabilityImpact":"HIGH","baseScore":5.5,"baseSeverity":"MEDIUM"},"exploitabilityScore":1.8,"impactScore":3.6},"baseMetricV2":{"cvssV2":{"version":"2.0","vectorString":"AV:L/AC:L/Au:N/C:N/I:N/A:P","accessVector":"LOCAL","accessComplexity":"LOW","authentication":"NONE","confidentialityImpact":"NONE","integrityImpact":"NONE","availabilityImpact":"PARTIAL","baseScore":2.1},"severity":"LOW","exploitabilityScore":3.9,"impactScore":2.9,"acInsufInfo":false,"obtainAllPrivilege":false,"obtainUserPrivilege":false,"obtainOtherPrivilege":false,"userInteractionRequired":false}},"configurations":{"CVE_data_version":"4.0","nodes":[{"operator":"OR","children":[],"cpe_match":[{"vulnerable":true,"cpe23Uri":"cpe:2.3:a:google:tensorflow:*:*:*:*:*:*:*:*","versionStartIncluding":"2.4.0","versionEndExcluding":"2.4.2","cpe_name":[]},{"vulnerable":true,"cpe23Uri":"cpe:2.3:a:google:tensorflow:*:*:*:*:*:*:*:*","versionStartIncluding":"2.3.0","versionEndExcluding":"2.3.3","cpe_name":[]}]}]}},"legacy_mitre":{"record":{"CveYear":"2021","CveId":"29521","Ordinal":"204725","Title":"CVE-2021-29521","CVE":"CVE-2021-29521","Year":"2021"},"notes":[{"CveYear":"2021","CveId":"29521","Ordinal":"1","NoteData":"TensorFlow is an end-to-end open source platform for machine learning. Specifying a negative dense shape in `tf.raw_ops.SparseCountSparseOutput` results in a segmentation fault being thrown out from the standard library as `std::vector` invariants are broken. This is because the implementation(https://github.com/tensorflow/tensorflow/blob/8f7b60ee8c0206a2c99802e3a4d1bb55d2bc0624/tensorflow/core/kernels/count_ops.cc#L199-L213) assumes the first element of the dense shape is always positive and uses it to initialize a `BatchedMap<T>` (i.e., `std::vector<absl::flat_hash_map<int64,T>>`(https://github.com/tensorflow/tensorflow/blob/8f7b60ee8c0206a2c99802e3a4d1bb55d2bc0624/tensorflow/core/kernels/count_ops.cc#L27)) data structure. If the `shape` tensor has more than one element, `num_batches` is the first value in `shape`. Ensuring that the `dense_shape` argument is a valid tensor shape (that is, all elements are non-negative) solves this issue. The fix will be included in TensorFlow 2.5.0. We will also cherrypick this commit on TensorFlow 2.4.2 and TensorFlow 2.3.3.","Type":"Description","Title":null},{"CveYear":"2021","CveId":"29521","Ordinal":"2","NoteData":"2021-05-14","Type":"Other","Title":"Published"},{"CveYear":"2021","CveId":"29521","Ordinal":"3","NoteData":"2021-05-14","Type":"Other","Title":"Modified"}]}}}