{"id":"CVE-2021-37639","aliases":["BIT-tensorflow-2021-37639","EUVD-2021-0365","GHSA-gh6x-4whr-2qv4","PYSEC-2021-261","PYSEC-2021-552","PYSEC-2021-750","openSUSE-SU-2022:10014-1","openSUSE-SU-2024:12116-1"],"status":"published","title":"Null pointer dereference and heap OOB read in TensorFlow","description":"TensorFlow is an end-to-end open source platform for machine learning. When restoring tensors via raw APIs, if the tensor name is not provided, TensorFlow can be tricked into dereferencing a null pointer. Alternatively, attackers can read memory outside the bounds of heap allocated data by providing some tensor names but not enough for a successful restoration. The [implementation](https://github.com/tensorflow/tensorflow/blob/47a06f40411a69c99f381495f490536972152ac0/tensorflow/core/kernels/save_restore_tensor.cc#L158-L159) retrieves the tensor list corresponding to the `tensor_name` user controlled input and immediately retrieves the tensor at the restoration index (controlled via `preferred_shard` argument). This occurs without validating that the provided list has enough values. If the list is empty this results in dereferencing a null pointer (undefined behavior). If, however, the list has some elements, if the restoration index is outside the bounds this results in heap OOB read. We have patched the issue in GitHub commit 9e82dce6e6bd1f36a57e08fa85af213e2b2f2622. The fix will be included in TensorFlow 2.6.0. We will also cherrypick this commit on TensorFlow 2.5.1, TensorFlow 2.4.3, and TensorFlow 2.3.4, as these are also affected and still in supported range.","severity":{"version":"3.1","score":8.4,"vector":"CVSS:3.1/AV:L/AC:L/PR:N/UI:N/S:U/C:H/I:H/A:H","level":"high","assessor":"cna","from":"cve"},"epss":{"day":"2026-11-10","score":0.00173,"percentile":0.06867,"model":"v2026.06.15"},"facts":{"in_kev":false,"known_exploited":false,"public_exploit":false,"fix_available":true},"cwes":["CWE-125","CWE-476"],"published":"2021-08-12T18:10:15Z","modified":"2026-10-20T14:28:11Z","fixes":[{"ecosystem":"PyPI","package":"tensorflow","fixed_in":["2.3.4","2.4.3","2.5.1"],"sources":["ghsa","osv"]},{"ecosystem":"PyPI","package":"tensorflow-cpu","fixed_in":["2.3.4","2.4.3","2.5.1"],"sources":["ghsa","osv"]},{"ecosystem":"PyPI","package":"tensorflow-gpu","fixed_in":["2.3.4","2.4.3","2.5.1"],"sources":["ghsa","osv"]},{"ecosystem":"Bitnami","package":"tensorflow","fixed_in":["2.3.4","2.4.3","2.5.1"],"sources":["osv"]}],"affected":[{"source":"ghsa","ecosystem":"PyPI","package":"tensorflow","ranges":[{"type":"ECOSYSTEM","events":[{"introduced":"0"},{"fixed":"2.3.4"}]}]},{"source":"ghsa","ecosystem":"PyPI","package":"tensorflow","ranges":[{"type":"ECOSYSTEM","events":[{"introduced":"2.4.0"},{"fixed":"2.4.3"}]}]},{"source":"ghsa","ecosystem":"PyPI","package":"tensorflow","ranges":[{"type":"ECOSYSTEM","events":[{"introduced":"2.5.0"},{"last_affected":"2.5.0"},{"fixed":"2.5.1"}]}],"versions":["2.5.0"]},{"source":"ghsa","ecosystem":"PyPI","package":"tensorflow-cpu","ranges":[{"type":"ECOSYSTEM","events":[{"introduced":"0"},{"fixed":"2.3.4"}]}]},{"source":"ghsa","ecosystem":"PyPI","package":"tensorflow-cpu","ranges":[{"type":"ECOSYSTEM","events":[{"introduced":"2.4.0"},{"fixed":"2.4.3"}]}]},{"source":"ghsa","ecosystem":"PyPI","package":"tensorflow-cpu","ranges":[{"type":"ECOSYSTEM","events":[{"introduced":"2.5.0"},{"last_affected":"2.5.0"},{"fixed":"2.5.1"}]}],"versions":["2.5.0"]},{"source":"ghsa","ecosystem":"PyPI","package":"tensorflow-gpu","ranges":[{"type":"ECOSYSTEM","events":[{"introduced":"0"},{"fixed":"2.3.4"}]}]},{"source":"ghsa","ecosystem":"PyPI","package":"tensorflow-gpu","ranges":[{"type":"ECOSYSTEM","events":[{"introduced":"2.4.0"},{"fixed":"2.4.3"}]}]},{"source":"ghsa","ecosystem":"PyPI","package":"tensorflow-gpu","ranges":[{"type":"ECOSYSTEM","events":[{"introduced":"2.5.0"},{"last_affected":"2.5.0"},{"fixed":"2.5.1"}]}],"versions":["2.5.0"]},{"source":"osv","ecosystem":"Bitnami","package":"tensorflow","ranges":[{"type":"SEMVER","events":[{"introduced":"2.3.0"},{"fixed":"2.3.4"},{"introduced":"2.4.0"},{"fixed":"2.4.3"},{"introduced":"2.5.0"},{"fixed":"2.5.1"}]}]},{"source":"osv","ecosystem":"PyPI","package":"tensorflow","ranges":[{"type":"ECOSYSTEM","events":[{"introduced":"0"},{"fixed":"2.3.4"}]}],"versions":["0.12.0","0.12.1","1.0.0","1.0.1","1.1.0","1.10.0","1.10.1","1.11.0","1.12.0","1.12.2","1.12.3","1.13.1","1.13.2","1.14.0","1.15.0","1.15.2","1.15.3","1.15.4","1.15.5","1.2.0","1.2.1","1.3.0","1.4.0","1.4.1","1.5.0","1.5.1","1.6.0","1.7.0","1.7.1","1.8.0","1.9.0","2.0.0","2.0.1","2.0.2","2.0.3","2.0.4","2.1.0","2.1.1","2.1.2","2.1.3","2.1.4","2.2.0","2.2.1","2.2.2","2.2.3","2.3.0","2.3.1","2.3.2","2.3.3"]},{"source":"osv","ecosystem":"PyPI","package":"tensorflow","ranges":[{"type":"ECOSYSTEM","events":[{"introduced":"2.4.0"},{"fixed":"2.4.3"}]}],"versions":["2.4.0","2.4.1","2.4.2"]},{"source":"osv","ecosystem":"PyPI","package":"tensorflow","ranges":[{"type":"ECOSYSTEM","events":[{"introduced":"2.5.0"},{"fixed":"2.5.1"}]}],"versions":["2.5.0"]},{"source":"osv","ecosystem":"PyPI","package":"tensorflow-cpu","ranges":[{"type":"ECOSYSTEM","events":[{"introduced":"0"},{"fixed":"2.3.4"}]}],"versions":["1.15.0","2.1.0","2.1.1","2.1.2","2.1.3","2.1.4","2.2.0","2.2.1","2.2.2","2.2.3","2.3.0","2.3.1","2.3.2","2.3.3"]},{"source":"osv","ecosystem":"PyPI","package":"tensorflow-cpu","ranges":[{"type":"ECOSYSTEM","events":[{"introduced":"2.4.0"},{"fixed":"2.4.3"}]}],"versions":["2.4.0","2.4.1","2.4.2"]},{"source":"osv","ecosystem":"PyPI","package":"tensorflow-cpu","ranges":[{"type":"ECOSYSTEM","events":[{"introduced":"2.5.0"},{"fixed":"2.5.1"}]}],"versions":["2.5.0"]},{"source":"osv","ecosystem":"PyPI","package":"tensorflow-gpu","ranges":[{"type":"ECOSYSTEM","events":[{"introduced":"0"},{"fixed":"2.3.4"}]}],"versions":["0.12.0","0.12.1","1.0.0","1.0.1","1.1.0","1.10.0","1.10.1","1.11.0","1.12.0","1.12.2","1.12.3","1.13.1","1.13.2","1.14.0","1.15.0","1.15.2","1.15.3","1.15.4","1.15.5","1.2.0","1.2.1","1.3.0","1.4.0","1.4.1","1.5.0","1.5.1","1.6.0","1.7.0","1.7.1","1.8.0","1.9.0","2.0.0","2.0.1","2.0.2","2.0.3","2.0.4","2.1.0","2.1.1","2.1.2","2.1.3","2.1.4","2.2.0","2.2.1","2.2.2","2.2.3","2.3.0","2.3.1","2.3.2","2.3.3"]},{"source":"osv","ecosystem":"PyPI","package":"tensorflow-gpu","ranges":[{"type":"ECOSYSTEM","events":[{"introduced":"2.4.0"},{"fixed":"2.4.3"}]}],"versions":["2.4.0","2.4.1","2.4.2"]},{"source":"osv","ecosystem":"PyPI","package":"tensorflow-gpu","ranges":[{"type":"ECOSYSTEM","events":[{"introduced":"2.5.0"},{"fixed":"2.5.1"}]}],"versions":["2.5.0"]},{"source":"osv","ecosystem":"PyPI","package":"tensorflow","ranges":[{"type":"GIT","events":[{"introduced":"0"},{"fixed":"9e82dce6e6bd1f36a57e08fa85af213e2b2f2622"}]},{"type":"ECOSYSTEM","events":[{"introduced":"2.3.0"},{"fixed":"2.3.4"},{"introduced":"2.4.0"},{"fixed":"2.4.3"}]}],"versions":["2.3.0","2.3.1","2.3.2","2.3.3","2.4.0","2.4.1","2.4.2"]},{"source":"osv","ecosystem":"PyPI","package":"tensorflow-cpu","ranges":[{"type":"GIT","events":[{"introduced":"0"},{"fixed":"9e82dce6e6bd1f36a57e08fa85af213e2b2f2622"}]},{"type":"ECOSYSTEM","events":[{"introduced":"2.3.0"},{"fixed":"2.3.4"},{"introduced":"2.4.0"},{"fixed":"2.4.3"}]}],"versions":["2.3.0","2.3.1","2.3.2","2.3.3","2.4.0","2.4.1","2.4.2"]},{"source":"osv","ecosystem":"PyPI","package":"tensorflow-gpu","ranges":[{"type":"GIT","events":[{"introduced":"0"},{"fixed":"9e82dce6e6bd1f36a57e08fa85af213e2b2f2622"}]},{"type":"ECOSYSTEM","events":[{"introduced":"2.3.0"},{"fixed":"2.3.4"},{"introduced":"2.4.0"},{"fixed":"2.4.3"}]}],"versions":["2.3.0","2.3.1","2.3.2","2.3.3","2.4.0","2.4.1","2.4.2"]},{"source":"osv","ecosystem":"SUSE:Package Hub 15 SP3","package":"bazel-skylib1.0.3","ranges":[{"type":"ECOSYSTEM","events":[{"introduced":"0"},{"fixed":"1.0.3-bp153.2.1"}]}],"bundled":63},{"source":"osv","ecosystem":"SUSE:Package Hub 15 SP3","package":"bazel3.7","ranges":[{"type":"ECOSYSTEM","events":[{"introduced":"0"},{"fixed":"3.7.2-bp153.4.1"}]}],"bundled":63},{"source":"osv","ecosystem":"SUSE:Package Hub 15 SP3","package":"tensorflow2","ranges":[{"type":"ECOSYSTEM","events":[{"introduced":"0"},{"fixed":"2.6.0-bp153.2.3.1"}]}],"bundled":63},{"source":"osv","ecosystem":"SUSE:Package Hub 15 SP3","package":"tensorflow2-lite","ranges":[{"type":"ECOSYSTEM","events":[{"introduced":"0"},{"fixed":"2.6.0-bp153.2.3.1"}]}],"bundled":63},{"source":"osv","ecosystem":"SUSE:Package Hub 15 SP3","package":"tensorflow2_2_6_0-gnu-hpc","ranges":[{"type":"ECOSYSTEM","events":[{"introduced":"0"},{"fixed":"2.6.0-bp153.2.3.1"}]}],"bundled":63},{"source":"osv","ecosystem":"SUSE:Package Hub 15 SP3","package":"tensorflow2_2_6_0-gnu-openmpi2-hpc","ranges":[{"type":"ECOSYSTEM","events":[{"introduced":"0"},{"fixed":"2.6.0-bp153.2.3.1"}]}],"bundled":63},{"source":"osv","ecosystem":"openSUSE:Leap 15.3","package":"bazel-skylib1.0.3","ranges":[{"type":"ECOSYSTEM","events":[{"introduced":"0"},{"fixed":"1.0.3-bp153.2.1"}]}],"bundled":63},{"source":"osv","ecosystem":"openSUSE:Leap 15.3","package":"bazel3.7","ranges":[{"type":"ECOSYSTEM","events":[{"introduced":"0"},{"fixed":"3.7.2-bp153.4.1"}]}],"bundled":63},{"source":"osv","ecosystem":"openSUSE:Leap 15.3","package":"tensorflow2","ranges":[{"type":"ECOSYSTEM","events":[{"introduced":"0"},{"fixed":"2.6.0-bp153.2.3.1"}]}],"bundled":63},{"source":"osv","ecosystem":"openSUSE:Leap 15.3","package":"tensorflow2-lite","ranges":[{"type":"ECOSYSTEM","events":[{"introduced":"0"},{"fixed":"2.6.0-bp153.2.3.1"}]}],"bundled":63},{"source":"osv","ecosystem":"openSUSE:Leap 15.3","package":"tensorflow2_2_6_0-gnu-hpc","ranges":[{"type":"ECOSYSTEM","events":[{"introduced":"0"},{"fixed":"2.6.0-bp153.2.3.1"}]}],"bundled":63},{"source":"osv","ecosystem":"openSUSE:Leap 15.3","package":"tensorflow2_2_6_0-gnu-openmpi2-hpc","ranges":[{"type":"ECOSYSTEM","events":[{"introduced":"0"},{"fixed":"2.6.0-bp153.2.3.1"}]}],"bundled":63}],"products":[{"vendor":"google","product":"tensorflow"},{"vendor":"tensorflow","product":"tensorflow"}],"references":[{"url":"https://github.com/tensorflow/tensorflow/security/advisories/GHSA-gh6x-4whr-2qv4","tags":["x_refsource_CONFIRM","x_transferred","Third Party Advisory","advisory","web"]},{"url":"https://github.com/tensorflow/tensorflow/commit/9e82dce6e6bd1f36a57e08fa85af213e2b2f2622","tags":["x_refsource_MISC","x_transferred","Patch","Third Party Advisory","advisory","fix","web"]},{"url":"https://nvd.nist.gov/vuln/detail/CVE-2021-37639","tags":["web","advisory"]},{"url":"https://github.com/pypa/advisory-database/tree/main/vulns/tensorflow-cpu/PYSEC-2021-552.yaml","tags":["web"]},{"url":"https://github.com/pypa/advisory-database/tree/main/vulns/tensorflow-gpu/PYSEC-2021-750.yaml","tags":["web"]},{"url":"https://github.com/pypa/advisory-database/tree/main/vulns/tensorflow/PYSEC-2021-261.yaml","tags":["web"]},{"url":"https://github.com/advisories/GHSA-gh6x-4whr-2qv4"},{"url":"https://github.com/tensorflow/tensorflow","tags":["package"]},{"url":"https://lists.opensuse.org/archives/list/security-announce@lists.opensuse.org/thread/U25ZU2T5T4LOLSIYIMGC5PLU4PQQMJE5/","tags":["advisory"]},{"url":"https://bugzilla.suse.com/1173128","tags":["report"]},{"url":"https://bugzilla.suse.com/1173314","tags":["report"]},{"url":"https://bugzilla.suse.com/1178287","tags":["report"]},{"url":"https://bugzilla.suse.com/1178564","tags":["report"]},{"url":"https://bugzilla.suse.com/1179455","tags":["report"]},{"url":"https://bugzilla.suse.com/1181864","tags":["report"]},{"url":"https://bugzilla.suse.com/1186860","tags":["report"]},{"url":"https://bugzilla.suse.com/1189423","tags":["report"]},{"url":"https://www.suse.com/security/cve/CVE-2020-26266","tags":["web"]},{"url":"https://www.suse.com/security/cve/CVE-2020-26267","tags":["web"]},{"url":"https://www.suse.com/security/cve/CVE-2020-26268","tags":["web"]},{"url":"https://www.suse.com/security/cve/CVE-2020-26270","tags":["web"]},{"url":"https://www.suse.com/security/cve/CVE-2020-26271","tags":["web"]},{"url":"https://www.suse.com/security/cve/CVE-2021-37635","tags":["web"]},{"url":"https://www.suse.com/security/cve/CVE-2021-37636","tags":["web"]},{"url":"https://www.suse.com/security/cve/CVE-2021-37637","tags":["web"]},{"url":"https://www.suse.com/security/cve/CVE-2021-37638","tags":["web"]},{"url":"https://www.suse.com/security/cve/CVE-2021-37639","tags":["web"]},{"url":"https://www.suse.com/security/cve/CVE-2021-37640","tags":["web"]},{"url":"https://www.suse.com/security/cve/CVE-2021-37641","tags":["web"]},{"url":"https://www.suse.com/security/cve/CVE-2021-37642","tags":["web"]},{"url":"https://www.suse.com/security/cve/CVE-2021-37643","tags":["web"]},{"url":"https://www.suse.com/security/cve/CVE-2021-37644","tags":["web"]},{"url":"https://www.suse.com/security/cve/CVE-2021-37645","tags":["web"]},{"url":"https://www.suse.com/security/cve/CVE-2021-37646","tags":["web"]},{"url":"https://www.suse.com/security/cve/CVE-2021-37647","tags":["web"]},{"url":"https://www.suse.com/security/cve/CVE-2021-37648","tags":["web"]},{"url":"https://www.suse.com/security/cve/CVE-2021-37649","tags":["web"]},{"url":"https://www.suse.com/security/cve/CVE-2021-37650","tags":["web"]},{"url":"https://www.suse.com/security/cve/CVE-2021-37651","tags":["web"]},{"url":"https://www.suse.com/security/cve/CVE-2021-37652","tags":["web"]},{"url":"https://www.suse.com/security/cve/CVE-2021-37653","tags":["web"]},{"url":"https://www.suse.com/security/cve/CVE-2021-37654","tags":["web"]},{"url":"https://www.suse.com/security/cve/CVE-2021-37655","tags":["web"]},{"url":"https://www.suse.com/security/cve/CVE-2021-37656","tags":["web"]},{"url":"https://www.suse.com/security/cve/CVE-2021-37657","tags":["web"]},{"url":"https://www.suse.com/security/cve/CVE-2021-37658","tags":["web"]},{"url":"https://www.suse.com/security/cve/CVE-2021-37659","tags":["web"]},{"url":"https://www.suse.com/security/cve/CVE-2021-37660","tags":["web"]},{"url":"https://www.suse.com/security/cve/CVE-2021-37661","tags":["web"]},{"url":"https://www.suse.com/security/cve/CVE-2021-37662","tags":["web"]},{"url":"https://www.suse.com/security/cve/CVE-2021-37663","tags":["web"]},{"url":"https://www.suse.com/security/cve/CVE-2021-37664","tags":["web"]},{"url":"https://www.suse.com/security/cve/CVE-2021-37665","tags":["web"]},{"url":"https://www.suse.com/security/cve/CVE-2021-37666","tags":["web"]},{"url":"https://www.suse.com/security/cve/CVE-2021-37667","tags":["web"]},{"url":"https://www.suse.com/security/cve/CVE-2021-37668","tags":["web"]},{"url":"https://www.suse.com/security/cve/CVE-2021-37669","tags":["web"]},{"url":"https://www.suse.com/security/cve/CVE-2021-37670","tags":["web"]},{"url":"https://www.suse.com/security/cve/CVE-2021-37671","tags":["web"]},{"url":"https://www.suse.com/security/cve/CVE-2021-37672","tags":["web"]},{"url":"https://www.suse.com/security/cve/CVE-2021-37673","tags":["web"]},{"url":"https://www.suse.com/security/cve/CVE-2021-37674","tags":["web"]},{"url":"https://www.suse.com/security/cve/CVE-2021-37675","tags":["web"]},{"url":"https://www.suse.com/security/cve/CVE-2021-37676","tags":["web"]},{"url":"https://www.suse.com/security/cve/CVE-2021-37677","tags":["web"]},{"url":"https://www.suse.com/security/cve/CVE-2021-37678","tags":["web"]},{"url":"https://www.suse.com/security/cve/CVE-2021-37679","tags":["web"]},{"url":"https://www.suse.com/security/cve/CVE-2021-37680","tags":["web"]},{"url":"https://www.suse.com/security/cve/CVE-2021-37681","tags":["web"]},{"url":"https://www.suse.com/security/cve/CVE-2021-37682","tags":["web"]},{"url":"https://www.suse.com/security/cve/CVE-2021-37683","tags":["web"]},{"url":"https://www.suse.com/security/cve/CVE-2021-37684","tags":["web"]},{"url":"https://www.suse.com/security/cve/CVE-2021-37685","tags":["web"]},{"url":"https://www.suse.com/security/cve/CVE-2021-37686","tags":["web"]},{"url":"https://www.suse.com/security/cve/CVE-2021-37687","tags":["web"]},{"url":"https://www.suse.com/security/cve/CVE-2021-37688","tags":["web"]},{"url":"https://www.suse.com/security/cve/CVE-2021-37689","tags":["web"]},{"url":"https://www.suse.com/security/cve/CVE-2021-37690","tags":["web"]},{"url":"https://www.suse.com/security/cve/CVE-2021-37691","tags":["web"]},{"url":"https://www.suse.com/security/cve/CVE-2021-37692","tags":["web"]},{"url":"","tags":["advisory"]},{"url":"https://www.suse.com/security/cve/CVE-2020-10531","tags":["web"]},{"url":"https://www.suse.com/security/cve/CVE-2020-15190","tags":["web"]},{"url":"https://www.suse.com/security/cve/CVE-2020-15191","tags":["web"]},{"url":"https://www.suse.com/security/cve/CVE-2020-15194","tags":["web"]},{"url":"https://www.suse.com/security/cve/CVE-2020-15202","tags":["web"]},{"url":"https://www.suse.com/security/cve/CVE-2020-15203","tags":["web"]},{"url":"https://www.suse.com/security/cve/CVE-2020-15204","tags":["web"]},{"url":"https://www.suse.com/security/cve/CVE-2020-15205","tags":["web"]},{"url":"https://www.suse.com/security/cve/CVE-2020-15206","tags":["web"]},{"url":"https://www.suse.com/security/cve/CVE-2020-15207","tags":["web"]},{"url":"https://www.suse.com/security/cve/CVE-2020-15208","tags":["web"]},{"url":"https://www.suse.com/security/cve/CVE-2020-15209","tags":["web"]},{"url":"https://www.suse.com/security/cve/CVE-2021-22922","tags":["web"]},{"url":"https://www.suse.com/security/cve/CVE-2021-22923","tags":["web"]},{"url":"https://www.suse.com/security/cve/CVE-2021-22926","tags":["web"]},{"url":"https://www.suse.com/security/cve/CVE-2021-41195","tags":["web"]},{"url":"https://www.suse.com/security/cve/CVE-2021-41196","tags":["web"]},{"url":"https://www.suse.com/security/cve/CVE-2021-41197","tags":["web"]},{"url":"https://www.suse.com/security/cve/CVE-2021-41198","tags":["web"]},{"url":"https://www.suse.com/security/cve/CVE-2021-41199","tags":["web"]},{"url":"https://www.suse.com/security/cve/CVE-2021-41200","tags":["web"]},{"url":"https://www.suse.com/security/cve/CVE-2021-41201","tags":["web"]},{"url":"https://www.suse.com/security/cve/CVE-2021-41202","tags":["web"]},{"url":"https://www.suse.com/security/cve/CVE-2021-41203","tags":["web"]},{"url":"https://www.suse.com/security/cve/CVE-2021-41204","tags":["web"]},{"url":"https://www.suse.com/security/cve/CVE-2021-41205","tags":["web"]},{"url":"https://www.suse.com/security/cve/CVE-2021-41206","tags":["web"]},{"url":"https://www.suse.com/security/cve/CVE-2021-41207","tags":["web"]},{"url":"https://www.suse.com/security/cve/CVE-2021-41208","tags":["web"]},{"url":"https://www.suse.com/security/cve/CVE-2021-41209","tags":["web"]},{"url":"https://www.suse.com/security/cve/CVE-2021-41210","tags":["web"]},{"url":"https://www.suse.com/security/cve/CVE-2021-41211","tags":["web"]},{"url":"https://www.suse.com/security/cve/CVE-2021-41212","tags":["web"]},{"url":"https://www.suse.com/security/cve/CVE-2021-41213","tags":["web"]},{"url":"https://www.suse.com/security/cve/CVE-2021-41214","tags":["web"]},{"url":"https://www.suse.com/security/cve/CVE-2021-41215","tags":["web"]},{"url":"https://www.suse.com/security/cve/CVE-2021-41216","tags":["web"]},{"url":"https://www.suse.com/security/cve/CVE-2021-41217","tags":["web"]},{"url":"https://www.suse.com/security/cve/CVE-2021-41218","tags":["web"]},{"url":"https://www.suse.com/security/cve/CVE-2021-41219","tags":["web"]},{"url":"https://www.suse.com/security/cve/CVE-2021-41220","tags":["web"]},{"url":"https://www.suse.com/security/cve/CVE-2021-41221","tags":["web"]},{"url":"https://www.suse.com/security/cve/CVE-2021-41222","tags":["web"]},{"url":"https://www.suse.com/security/cve/CVE-2021-41223","tags":["web"]},{"url":"https://www.suse.com/security/cve/CVE-2021-41224","tags":["web"]},{"url":"https://www.suse.com/security/cve/CVE-2021-41225","tags":["web"]},{"url":"https://www.suse.com/security/cve/CVE-2021-41226","tags":["web"]},{"url":"https://www.suse.com/security/cve/CVE-2021-41227","tags":["web"]},{"url":"https://www.suse.com/security/cve/CVE-2021-41228","tags":["web"]},{"url":"https://www.suse.com/security/cve/CVE-2022-21725","tags":["web"]},{"url":"https://www.suse.com/security/cve/CVE-2022-21726","tags":["web"]},{"url":"https://www.suse.com/security/cve/CVE-2022-21727","tags":["web"]},{"url":"https://www.suse.com/security/cve/CVE-2022-21728","tags":["web"]},{"url":"https://www.suse.com/security/cve/CVE-2022-21729","tags":["web"]},{"url":"https://www.suse.com/security/cve/CVE-2022-21730","tags":["web"]},{"url":"https://www.suse.com/security/cve/CVE-2022-21731","tags":["web"]},{"url":"https://www.suse.com/security/cve/CVE-2022-21732","tags":["web"]},{"url":"https://www.suse.com/security/cve/CVE-2022-21733","tags":["web"]},{"url":"https://www.suse.com/security/cve/CVE-2022-21734","tags":["web"]},{"url":"https://www.suse.com/security/cve/CVE-2022-21735","tags":["web"]},{"url":"https://www.suse.com/security/cve/CVE-2022-21736","tags":["web"]},{"url":"https://www.suse.com/security/cve/CVE-2022-21737","tags":["web"]},{"url":"https://www.suse.com/security/cve/CVE-2022-21738","tags":["web"]},{"url":"https://www.suse.com/security/cve/CVE-2022-21739","tags":["web"]},{"url":"https://www.suse.com/security/cve/CVE-2022-21740","tags":["web"]},{"url":"https://www.suse.com/security/cve/CVE-2022-21741","tags":["web"]},{"url":"https://www.suse.com/security/cve/CVE-2022-23557","tags":["web"]},{"url":"https://www.suse.com/security/cve/CVE-2022-23558","tags":["web"]},{"url":"https://www.suse.com/security/cve/CVE-2022-23559","tags":["web"]},{"url":"https://www.suse.com/security/cve/CVE-2022-23560","tags":["web"]},{"url":"https://www.suse.com/security/cve/CVE-2022-23561","tags":["web"]},{"url":"https://www.suse.com/security/cve/CVE-2022-23562","tags":["web"]},{"url":"https://www.suse.com/security/cve/CVE-2022-23563","tags":["web"]},{"url":"https://www.suse.com/security/cve/CVE-2022-23564","tags":["web"]},{"url":"https://www.suse.com/security/cve/CVE-2022-23565","tags":["web"]},{"url":"https://www.suse.com/security/cve/CVE-2022-23566","tags":["web"]},{"url":"https://www.suse.com/security/cve/CVE-2022-23567","tags":["web"]},{"url":"https://www.suse.com/security/cve/CVE-2022-23568","tags":["web"]},{"url":"https://www.suse.com/security/cve/CVE-2022-23569","tags":["web"]},{"url":"https://www.suse.com/security/cve/CVE-2022-23570","tags":["web"]},{"url":"https://www.suse.com/security/cve/CVE-2022-23571","tags":["web"]},{"url":"https://www.suse.com/security/cve/CVE-2022-23572","tags":["web"]},{"url":"https://www.suse.com/security/cve/CVE-2022-23573","tags":["web"]},{"url":"https://www.suse.com/security/cve/CVE-2022-23574","tags":["web"]},{"url":"https://www.suse.com/security/cve/CVE-2022-23575","tags":["web"]},{"url":"https://www.suse.com/security/cve/CVE-2022-23576","tags":["web"]},{"url":"https://www.suse.com/security/cve/CVE-2022-23577","tags":["web"]},{"url":"https://www.suse.com/security/cve/CVE-2022-23578","tags":["web"]},{"url":"https://www.suse.com/security/cve/CVE-2022-23579","tags":["web"]},{"url":"https://www.suse.com/security/cve/CVE-2022-23580","tags":["web"]},{"url":"https://www.suse.com/security/cve/CVE-2022-23581","tags":["web"]},{"url":"https://www.suse.com/security/cve/CVE-2022-23582","tags":["web"]},{"url":"https://www.suse.com/security/cve/CVE-2022-23583","tags":["web"]},{"url":"https://www.suse.com/security/cve/CVE-2022-23584","tags":["web"]},{"url":"https://www.suse.com/security/cve/CVE-2022-23585","tags":["web"]},{"url":"https://www.suse.com/security/cve/CVE-2022-23586","tags":["web"]},{"url":"https://www.suse.com/security/cve/CVE-2022-23587","tags":["web"]},{"url":"https://www.suse.com/security/cve/CVE-2022-23588","tags":["web"]},{"url":"https://www.suse.com/security/cve/CVE-2022-23589","tags":["web"]},{"url":"https://www.suse.com/security/cve/CVE-2022-23590","tags":["web"]},{"url":"https://www.suse.com/security/cve/CVE-2022-23591","tags":["web"]},{"url":"https://www.suse.com/security/cve/CVE-2022-23594","tags":["web"]},{"url":"https://www.suse.com/security/cve/CVE-2022-23595","tags":["web"]}],"sources":{"cve":[{"id":"CVE-2021-37639","title":"Null pointer dereference and heap OOB read in TensorFlow","description":"TensorFlow is an end-to-end open source platform for machine learning. When restoring tensors via raw APIs, if the tensor name is not provided, TensorFlow can be tricked into dereferencing a null pointer. Alternatively, attackers can read memory outside the bounds of heap allocated data by providing some tensor names but not enough for a successful restoration. The [implementation](https://github.com/tensorflow/tensorflow/blob/47a06f40411a69c99f381495f490536972152ac0/tensorflow/core/kernels/save_restore_tensor.cc#L158-L159) retrieves the tensor list corresponding to the `tensor_name` user controlled input and immediately retrieves the tensor at the restoration index (controlled via `preferred_shard` argument). This occurs without validating that the provided list has enough values. If the list is empty this results in dereferencing a null pointer (undefined behavior). If, however, the list has some elements, if the restoration index is outside the bounds this results in heap OOB read. We have patched the issue in GitHub commit 9e82dce6e6bd1f36a57e08fa85af213e2b2f2622. The fix will be included in TensorFlow 2.6.0. We will also cherrypick this commit on TensorFlow 2.5.1, TensorFlow 2.4.3, and TensorFlow 2.3.4, as these are also affected and still in supported range.","status":"published","published":"2021-08-12T18:10:15Z","modified":"2024-08-04T01:23:01.249Z","cvss":[{"version":"3.1","score":8.4,"vector":"CVSS:3.1/AV:L/AC:L/PR:N/UI:N/S:U/C:H/I:H/A:H","level":"high","assessor":"cna"}],"cwes":["CWE-476"],"updated_at":"2026-11-02T14:32:58.028461Z"}],"euvd":[{"id":"EUVD-2021-0365","description":"TensorFlow is an end-to-end open source platform for machine learning. When restoring tensors via raw APIs, if the tensor name is not provided, TensorFlow can be tricked into dereferencing a null pointer. Alternatively, attackers can read memory outside the bounds of heap allocated data by providing some tensor names but not enough for a successful restoration. The [implementation](https://github.com/tensorflow/tensorflow/blob/47a06f40411a69c99f381495f490536972152ac0/tensorflow/core/kernels/save_restore_tensor.cc#L158-L159) retrieves the tensor list corresponding to the `tensor_name` user controlled input and immediately retrieves the tensor at the restoration index (controlled via `preferred_shard` argument). This occurs without validating that the provided list has enough values. If the list is empty this results in dereferencing a null pointer (undefined behavior). If, however, the list has some elements, if the restoration index is outside the bounds this results in heap OOB read. We have patched the issue in GitHub commit 9e82dce6e6bd1f36a57e08fa85af213e2b2f2622. The fix will be included in TensorFlow 2.6.0. We will also cherrypick this commit on TensorFlow 2.5.1, TensorFlow 2.4.3, and TensorFlow 2.3.4, as these are also affected and still in supported range.","status":"published","published":"2021-08-12T18:10:15Z","modified":"2024-08-04T01:23:01Z","cvss":[{"version":"3.1","score":8.4,"vector":"CVSS:3.1/AV:L/AC:L/PR:N/UI:N/S:U/C:H/I:H/A:H","level":"high"}],"updated_at":"2026-11-03T03:24:27.254804Z"}],"ghsa":[{"id":"GHSA-gh6x-4whr-2qv4","title":"Null pointer dereference and heap OOB read in operations restoring tensors","description":"### Impact\nWhen restoring tensors via raw APIs, if the tensor name is not provided, TensorFlow can be tricked into dereferencing a null pointer:\n\n```python\nimport tensorflow as tf\n\ntf.raw_ops.Restore(\n  file_pattern=['/tmp'],\n  tensor_name=[], \n  default_value=21,\n  dt=tf.int,\n  preferred_shard=1)\n```\n  \nThe same undefined behavior can be triggered by `tf.raw_ops.RestoreSlice`:\n  \n```python\nimport tensorflow as tf\n\ntf.raw_ops.RestoreSlice(\n  file_pattern=['/tmp'],\n  tensor_name=[], \n  shape_and_slice='2',\n  dt=inp.array([tf.int]),\n  preferred_shard=1)\n```\n\nAlternatively, attackers can read memory outside the bounds of heap allocated data by providing some tensor names but not enough for a successful restoration:\n\n```python\nimport tensorflow as tf\n\ntf.raw_ops.Restore(\n  file_pattern=['/tmp'],\n  tensor_name=['x'], \n  default_value=21,\n  dt=tf.int,\n  preferred_shard=42)\n```\n  \nThe [implementation](https://github.com/tensorflow/tensorflow/blob/47a06f40411a69c99f381495f490536972152ac0/tensorflow/core/kernels/save_restore_tensor.cc#L158-L159) retrieves the tensor list corresponding to the `tensor_name` user controlled input and immediately retrieves the tensor at the restoration index (controlled via `preferred_shard` argument). This occurs without validating that the provided list has enough values.\n\nIf the list is empty this results in dereferencing a null pointer (undefined behavior). If, however, the list has some elements, if the restoration index is outside the bounds this results in heap OOB read.\n\n### Patches \nWe have patched the issue in GitHub commit [9e82dce6e6bd1f36a57e08fa85af213e2b2f2622](https://github.com/tensorflow/tensorflow/commit/9e82dce6e6bd1f36a57e08fa85af213e2b2f2622).\n\nThe fix will be included in TensorFlow 2.6.0. We will also cherrypick this commit on TensorFlow 2.5.1, TensorFlow 2.4.3, and TensorFlow 2.3.4, as these are also affected and still in supported range.\n\n### For more information \nPlease consult [our security guide](https://github.com/tensorflow/tensorflow/blob/master/SECURITY.md) for more information regarding the security model and how to contact us with issues and questions.\n\n### Attribution\nThis vulnerability has been reported by members of the Aivul Team from Qihoo 360.","status":"published","published":"2021-08-25T14:44:05Z","modified":"2024-11-13T16:33:03Z","cvss":[{"version":"3.1","score":8.4,"vector":"CVSS:3.1/AV:L/AC:L/PR:N/UI:N/S:U/C:H/I:H/A:H","level":"high"},{"version":"4.0","score":8.6,"vector":"CVSS:4.0/AV:L/AC:L/AT:N/PR:N/UI:N/VC:H/VI:H/VA:H/SC:N/SI:N/SA:N","level":"high"}],"cwes":["CWE-125","CWE-476"],"updated_at":"2026-11-02T14:38:17.357217Z"}],"nvd":[{"id":"CVE-2021-37639","description":"TensorFlow is an end-to-end open source platform for machine learning. When restoring tensors via raw APIs, if the tensor name is not provided, TensorFlow can be tricked into dereferencing a null pointer. Alternatively, attackers can read memory outside the bounds of heap allocated data by providing some tensor names but not enough for a successful restoration. The [implementation](https://github.com/tensorflow/tensorflow/blob/47a06f40411a69c99f381495f490536972152ac0/tensorflow/core/kernels/save_restore_tensor.cc#L158-L159) retrieves the tensor list corresponding to the `tensor_name` user controlled input and immediately retrieves the tensor at the restoration index (controlled via `preferred_shard` argument). This occurs without validating that the provided list has enough values. If the list is empty this results in dereferencing a null pointer (undefined behavior). If, however, the list has some elements, if the restoration index is outside the bounds this results in heap OOB read. We have patched the issue in GitHub commit 9e82dce6e6bd1f36a57e08fa85af213e2b2f2622. The fix will be included in TensorFlow 2.6.0. We will also cherrypick this commit on TensorFlow 2.5.1, TensorFlow 2.4.3, and TensorFlow 2.3.4, as these are also affected and still in supported range.","status":"published","published":"2021-08-12T19:15:08.707Z","modified":"2026-06-17T04:00:53.017Z","cvss":[{"version":"3.1","score":8.4,"vector":"CVSS:3.1/AV:L/AC:L/PR:N/UI:N/S:U/C:H/I:H/A:H","level":"high","assessor":"secondary"},{"version":"3.1","score":7.8,"vector":"CVSS:3.1/AV:L/AC:L/PR:L/UI:N/S:U/C:H/I:H/A:H","level":"high","assessor":"primary"},{"version":"2.0","score":4.6,"vector":"AV:L/AC:L/Au:N/C:P/I:P/A:P","level":"medium","assessor":"primary"}],"cwes":["CWE-476","CWE-125"],"updated_at":"2026-11-02T14:06:56.943671Z"}],"osv":[{"id":"CVE-2021-37639","description":"TensorFlow is an end-to-end open source platform for machine learning. When restoring tensors via raw APIs, if the tensor name is not provided, TensorFlow can be tricked into dereferencing a null pointer. Alternatively, attackers can read memory outside the bounds of heap allocated data by providing some tensor names but not enough for a successful restoration. The [implementation](https://github.com/tensorflow/tensorflow/blob/47a06f40411a69c99f381495f490536972152ac0/tensorflow/core/kernels/save_restore_tensor.cc#L158-L159) retrieves the tensor list corresponding to the `tensor_name` user controlled input and immediately retrieves the tensor at the restoration index (controlled via `preferred_shard` argument). This occurs without validating that the provided list has enough values. If the list is empty this results in dereferencing a null pointer (undefined behavior). If, however, the list has some elements, if the restoration index is outside the bounds this results in heap OOB read. We have patched the issue in GitHub commit 9e82dce6e6bd1f36a57e08fa85af213e2b2f2622. The fix will be included in TensorFlow 2.6.0. We will also cherrypick this commit on TensorFlow 2.5.1, TensorFlow 2.4.3, and TensorFlow 2.3.4, as these are also affected and still in supported range.","status":"published","published":"2021-08-12T19:15:08.707Z","modified":"2026-10-20T14:28:11.096249929Z","cvss":[{"version":"3.1","score":7.8,"vector":"CVSS:3.1/AV:L/AC:L/PR:L/UI:N/S:U/C:H/I:H/A:H","level":"high"}],"updated_at":"2026-11-02T15:44:20.434873Z"},{"id":"GHSA-gh6x-4whr-2qv4","title":"Null pointer dereference and heap OOB read in operations restoring tensors","description":"### Impact\nWhen restoring tensors via raw APIs, if the tensor name is not provided, TensorFlow can be tricked into dereferencing a null pointer:\n\n```python\nimport tensorflow as tf\n\ntf.raw_ops.Restore(\n  file_pattern=['/tmp'],\n  tensor_name=[], \n  default_value=21,\n  dt=tf.int,\n  preferred_shard=1)\n```\n  \nThe same undefined behavior can be triggered by `tf.raw_ops.RestoreSlice`:\n  \n```python\nimport tensorflow as tf\n\ntf.raw_ops.RestoreSlice(\n  file_pattern=['/tmp'],\n  tensor_name=[], \n  shape_and_slice='2',\n  dt=inp.array([tf.int]),\n  preferred_shard=1)\n```\n\nAlternatively, attackers can read memory outside the bounds of heap allocated data by providing some tensor names but not enough for a successful restoration:\n\n```python\nimport tensorflow as tf\n\ntf.raw_ops.Restore(\n  file_pattern=['/tmp'],\n  tensor_name=['x'], \n  default_value=21,\n  dt=tf.int,\n  preferred_shard=42)\n```\n  \nThe [implementation](https://github.com/tensorflow/tensorflow/blob/47a06f40411a69c99f381495f490536972152ac0/tensorflow/core/kernels/save_restore_tensor.cc#L158-L159) retrieves the tensor list corresponding to the `tensor_name` user controlled input and immediately retrieves the tensor at the restoration index (controlled via `preferred_shard` argument). This occurs without validating that the provided list has enough values.\n\nIf the list is empty this results in dereferencing a null pointer (undefined behavior). If, however, the list has some elements, if the restoration index is outside the bounds this results in heap OOB read.\n\n### Patches \nWe have patched the issue in GitHub commit [9e82dce6e6bd1f36a57e08fa85af213e2b2f2622](https://github.com/tensorflow/tensorflow/commit/9e82dce6e6bd1f36a57e08fa85af213e2b2f2622).\n\nThe fix will be included in TensorFlow 2.6.0. We will also cherrypick this commit on TensorFlow 2.5.1, TensorFlow 2.4.3, and TensorFlow 2.3.4, as these are also affected and still in supported range.\n\n### For more information \nPlease consult [our security guide](https://github.com/tensorflow/tensorflow/blob/master/SECURITY.md) for more information regarding the security model and how to contact us with issues and questions.\n\n### Attribution\nThis vulnerability has been reported by members of the Aivul Team from Qihoo 360.","status":"published","published":"2021-08-25T14:44:05Z","modified":"2026-07-08T06:29:26.668093854Z","cvss":[{"version":"3.1","score":8.4,"vector":"CVSS:3.1/AV:L/AC:L/PR:N/UI:N/S:U/C:H/I:H/A:H","level":"high"},{"version":"4.0","score":8.6,"vector":"CVSS:4.0/AV:L/AC:L/AT:N/PR:N/UI:N/VC:H/VI:H/VA:H/SC:N/SI:N/SA:N","level":"high"}],"cwes":["CWE-125","CWE-476"],"updated_at":"2026-11-02T15:46:47.822182Z"},{"id":"BIT-tensorflow-2021-37639","title":"Null pointer dereference and heap OOB read in TensorFlow","description":"TensorFlow is an end-to-end open source platform for machine learning. When restoring tensors via raw APIs, if the tensor name is not provided, TensorFlow can be tricked into dereferencing a null pointer. Alternatively, attackers can read memory outside the bounds of heap allocated data by providing some tensor names but not enough for a successful restoration. The [implementation](https://github.com/tensorflow/tensorflow/blob/47a06f40411a69c99f381495f490536972152ac0/tensorflow/core/kernels/save_restore_tensor.cc#L158-L159) retrieves the tensor list corresponding to the `tensor_name` user controlled input and immediately retrieves the tensor at the restoration index (controlled via `preferred_shard` argument). This occurs without validating that the provided list has enough values. If the list is empty this results in dereferencing a null pointer (undefined behavior). If, however, the list has some elements, if the restoration index is outside the bounds this results in heap OOB read. We have patched the issue in GitHub commit 9e82dce6e6bd1f36a57e08fa85af213e2b2f2622. The fix will be included in TensorFlow 2.6.0. We will also cherrypick this commit on TensorFlow 2.5.1, TensorFlow 2.4.3, and TensorFlow 2.3.4, as these are also affected and still in supported range.","status":"published","published":"2024-03-06T11:17:48.052Z","modified":"2026-09-08T08:48:22.996347845Z","updated_at":"2026-11-02T15:28:35.51001Z"},{"id":"openSUSE-SU-2022:10014-1","title":"Security update for tensorflow2","description":"\nThis update for tensorflow fixes the following issues:\n\n\nUpdate to TF2 2.6.0 which fixes multiple CVEs (boo#1189423).\n\n- Introduction of bazel6.3 and basel-skylib1.0.3 as build\n  dependencies.\n  The latter has been adapted to all a version in its package\n  name (if %set_ver_suffix is set to 1). This allows multiple\n  versions to exist for one product (not installed).\n  NOTE: basel-skylib1.0.3 does not exist in oS:Factory:\n  basel-skylib in oS:Factory - the base version - is 1.0.3.\n","status":"published","published":"2022-06-18T16:01:26Z","modified":"2026-02-04T02:20:34.480131Z","updated_at":"2026-11-02T16:58:32.653274Z"},{"id":"openSUSE-SU-2024:12116-1","title":"tensorflow-lite-2.9.1-1.1 on GA media","description":"These are all security issues fixed in the tensorflow-lite-2.9.1-1.1 package on the GA media of openSUSE Tumbleweed.","status":"published","published":"2024-06-15T00:00:00Z","modified":"2026-02-04T03:04:19.609756Z","updated_at":"2026-11-02T16:58:35.485255Z"},{"id":"PYSEC-2021-261","description":"TensorFlow is an end-to-end open source platform for machine learning. When restoring tensors via raw APIs, if the tensor name is not provided, TensorFlow can be tricked into dereferencing a null pointer. Alternatively, attackers can read memory outside the bounds of heap allocated data by providing some tensor names but not enough for a successful restoration. The [implementation](https://github.com/tensorflow/tensorflow/blob/47a06f40411a69c99f381495f490536972152ac0/tensorflow/core/kernels/save_restore_tensor.cc#L158-L159) retrieves the tensor list corresponding to the `tensor_name` user controlled input and immediately retrieves the tensor at the restoration index (controlled via `preferred_shard` argument). This occurs without validating that the provided list has enough values. If the list is empty this results in dereferencing a null pointer (undefined behavior). If, however, the list has some elements, if the restoration index is outside the bounds this results in heap OOB read. We have patched the issue in GitHub commit 9e82dce6e6bd1f36a57e08fa85af213e2b2f2622. The fix will be included in TensorFlow 2.6.0. We will also cherrypick this commit on TensorFlow 2.5.1, TensorFlow 2.4.3, and TensorFlow 2.3.4, as these are also affected and still in supported range.","status":"published","published":"2021-08-12T19:15:00Z","modified":"2023-12-06T01:01:21.482576Z","updated_at":"2026-11-02T15:52:52.682031Z"},{"id":"PYSEC-2021-552","description":"TensorFlow is an end-to-end open source platform for machine learning. When restoring tensors via raw APIs, if the tensor name is not provided, TensorFlow can be tricked into dereferencing a null pointer. Alternatively, attackers can read memory outside the bounds of heap allocated data by providing some tensor names but not enough for a successful restoration. The [implementation](https://github.com/tensorflow/tensorflow/blob/47a06f40411a69c99f381495f490536972152ac0/tensorflow/core/kernels/save_restore_tensor.cc#L158-L159) retrieves the tensor list corresponding to the `tensor_name` user controlled input and immediately retrieves the tensor at the restoration index (controlled via `preferred_shard` argument). This occurs without validating that the provided list has enough values. If the list is empty this results in dereferencing a null pointer (undefined behavior). If, however, the list has some elements, if the restoration index is outside the bounds this results in heap OOB read. We have patched the issue in GitHub commit 9e82dce6e6bd1f36a57e08fa85af213e2b2f2622. The fix will be included in TensorFlow 2.6.0. We will also cherrypick this commit on TensorFlow 2.5.1, TensorFlow 2.4.3, and TensorFlow 2.3.4, as these are also affected and still in supported range.","status":"published","published":"2021-08-12T19:15:00Z","modified":"2023-12-06T01:01:21.482576Z","updated_at":"2026-11-02T15:52:53.119977Z"},{"id":"PYSEC-2021-750","description":"TensorFlow is an end-to-end open source platform for machine learning. When restoring tensors via raw APIs, if the tensor name is not provided, TensorFlow can be tricked into dereferencing a null pointer. Alternatively, attackers can read memory outside the bounds of heap allocated data by providing some tensor names but not enough for a successful restoration. The [implementation](https://github.com/tensorflow/tensorflow/blob/47a06f40411a69c99f381495f490536972152ac0/tensorflow/core/kernels/save_restore_tensor.cc#L158-L159) retrieves the tensor list corresponding to the `tensor_name` user controlled input and immediately retrieves the tensor at the restoration index (controlled via `preferred_shard` argument). This occurs without validating that the provided list has enough values. If the list is empty this results in dereferencing a null pointer (undefined behavior). If, however, the list has some elements, if the restoration index is outside the bounds this results in heap OOB read. We have patched the issue in GitHub commit 9e82dce6e6bd1f36a57e08fa85af213e2b2f2622. The fix will be included in TensorFlow 2.6.0. We will also cherrypick this commit on TensorFlow 2.5.1, TensorFlow 2.4.3, and TensorFlow 2.3.4, as these are also affected and still in supported range.","status":"published","published":"2021-08-12T19:15:00Z","modified":"2023-12-06T01:01:21.482576Z","updated_at":"2026-11-02T15:52:53.119977Z"}]}}