← Vulnerability feed

Vulnerability record · CVE-2020-15211 · published 25 September 2020

CVE-2020-15211: Google tensorflow out-of-bounds read vulnerability

Google · Tensorflow

In TensorFlow Lite before versions 1.15.4, 2.0.3, 2.1.2, 2.2.1 and 2.3.1, saved models in the flatbuffer format use a double indexing scheme: a model has a set of subgraphs, each subgraph has a set of operators and each operator has a set of input/output tensors. The flatbuffer format uses indices for the tensors, indexing into an array of tensors that is owned by the subgraph. This results in a pattern of double array indexing when trying to get the data of each tensor. However, some operators can have some tensors be optional. To handle this scenario, the flatbuffer model uses a negative `-1` value as index for these tensors. This results in special casing during validation at model loading time. Unfortunately, this means that the `-1` index is a valid tensor index for any operator, including those that don't expect optional inputs and including for output tensors. Thus, this allows writing and reading from outside the bounds of heap allocated arrays, although only at a specific offset from the start of these arrays. This results in both read and write gadgets, albeit very limited in scope. The issue is patched in several commits (46d5b0852, 00302787b7, e11f5558, cd31fd0ce, 1970c21, and fff2c83), and is released in TensorFlow versions 1.15.4, 2.0.3, 2.1.2, 2.2.1, or 2.3.1. A potential workaround would be to add a custom `Verifier` to the model loading code to ensure that only operators which accept optional inputs use the `-1` special value and only for the tensors that they expect to be optional. Since this allow-list type approach is erro-prone, we advise upgrading to the patched code.

4.8 CVSS 3.1 Medium EPSS 0.92% · top 39.1% CWE-125 · Out-of-bounds readCWE-787 · Out-of-bounds write
4.8CVSS 3.1 base score, v2 5.8
0.92%EPSS exploitation probability, 30 days
NoNot in CISA KEV
2Affected product versions listed by NVD
18References, 2 tagged exploit
17 Jun 2026Last modified by NVD

Description

In TensorFlow Lite before versions 1.15.4, 2.0.3, 2.1.2, 2.2.1 and 2.3.1, saved models in the flatbuffer format use a double indexing scheme: a model has a set of subgraphs, each subgraph has a set of operators and each operator has a set of input/output tensors. The flatbuffer format uses indices for the tensors, indexing into an array of tensors that is owned by the subgraph. This results in a pattern of double array indexing when trying to get the data of each tensor. However, some operators can have some tensors be optional. To handle this scenario, the flatbuffer model uses a negative `-1` value as index for these tensors. This results in special casing during validation at model loading time. Unfortunately, this means that the `-1` index is a valid tensor index for any operator, including those that don't expect optional inputs and including for output tensors. Thus, this allows writing and reading from outside the bounds of heap allocated arrays, although only at a specific offset from the start of these arrays. This results in both read and write gadgets, albeit very limited in scope. The issue is patched in several commits (46d5b0852, 00302787b7, e11f5558, cd31fd0ce, 1970c21, and fff2c83), and is released in TensorFlow versions 1.15.4, 2.0.3, 2.1.2, 2.2.1, or 2.3.1. A potential workaround would be to add a custom `Verifier` to the model loading code to ensure that only operators which accept optional inputs use the `-1` special value and only for the tensors that they expect to be optional. Since this allow-list type approach is erro-prone, we advise upgrading to the patched code.

CVSS:3.1/AV:N/AC:H/PR:N/UI:N/S:U/C:L/I:L/A:N

Affected products

2 vulnerable configurations from NVD's CPE data, grouped by vendor and product.

References

LinkTags
http://lists.opensuse.org/opensuse-security-announce/2020-10/msg00065.html Mailing ListThird Party Advisory
https://github.com/tensorflow/tensorflow/commit/00302787b788c5ff04cb6f62aed5a74d936e86c0 PatchThird Party Advisory
https://github.com/tensorflow/tensorflow/commit/1970c2158b1ffa416d159d03c3370b9a462aee35 PatchThird Party Advisory
https://github.com/tensorflow/tensorflow/commit/46d5b0852528ddfd614ded79bccc75589f801bd9 PatchThird Party Advisory
https://github.com/tensorflow/tensorflow/commit/cd31fd0ce0449a9e0f83dcad08d6ed7f1d6bef3f PatchThird Party Advisory
https://github.com/tensorflow/tensorflow/commit/e11f55585f614645b360563072ffeb5c3eeff162 PatchThird Party Advisory
https://github.com/tensorflow/tensorflow/commit/fff2c8326280c07733828f990548979bdc893859 PatchThird Party Advisory
https://github.com/tensorflow/tensorflow/releases/tag/v2.3.1 Third Party Advisory
https://github.com/tensorflow/tensorflow/security/advisories/GHSA-cvpc-8phh-8f45 ExploitThird Party Advisory
http://lists.opensuse.org/opensuse-security-announce/2020-10/msg00065.html Mailing ListThird Party Advisory
https://github.com/tensorflow/tensorflow/commit/00302787b788c5ff04cb6f62aed5a74d936e86c0 PatchThird Party Advisory
https://github.com/tensorflow/tensorflow/commit/1970c2158b1ffa416d159d03c3370b9a462aee35 PatchThird Party Advisory
https://github.com/tensorflow/tensorflow/commit/46d5b0852528ddfd614ded79bccc75589f801bd9 PatchThird Party Advisory
https://github.com/tensorflow/tensorflow/commit/cd31fd0ce0449a9e0f83dcad08d6ed7f1d6bef3f PatchThird Party Advisory
https://github.com/tensorflow/tensorflow/commit/e11f55585f614645b360563072ffeb5c3eeff162 PatchThird Party Advisory
https://github.com/tensorflow/tensorflow/commit/fff2c8326280c07733828f990548979bdc893859 PatchThird Party Advisory
https://github.com/tensorflow/tensorflow/releases/tag/v2.3.1 Third Party Advisory
https://github.com/tensorflow/tensorflow/security/advisories/GHSA-cvpc-8phh-8f45 ExploitThird Party Advisory

Track CVE-2020-15211 inside VULONE

Watch it alongside the ransomware crews, C2 infrastructure and forum chatter that reference it, query it through the API and pull it into your SIEM over TAXII.

Start free Open in platform

Related vulnerabilities

Same products first, then exploited flaws of the same weakness class.

9.8CVE-2020-16846SaltStack Salt API shell injection via crafted web requestsSaltStack Salt through 3002 is vulnerable to OS command injection when the SSH client is enabled and crafted web requests are sent to the Salt API. T…KEVEPSS 100%analysed9.8CVE-2020-12641Roundcube Webmail OS command injection via image conversion path settingsRoundcube Webmail before 1.4.4 passes the im_convert_path and im_identify_path configuration settings to a shell without sanitization in rcube_image.…KEVEPSS 84%analysed9.8CVE-2020-11651SaltStack Salt master authentication bypass in ClearFuncsSaltStack Salt before 2019.2.4 and 3000 before 3000.2 fails to properly validate method calls in the salt-master ClearFuncs class, allowing remote un…KEVEPSS 97%analysed9.8CVE-2020-1938Apache Tomcat AJP connector file read and JSP execution flawApache Tomcat shipped an AJP Connector enabled by default that listened on all configured IP addresses, and Tomcat treats AJP connections as more tru…KEVEPSS 99%analysed9.8CVE-2016-3427Oracle Java SE JMX Improper Access Control VulnerabilityCVE-2016-3427 is an unspecified vulnerability in Oracle Java SE 6u113, 7u99, 8u77, Java SE Embedded 8u77, and JRockit R28.3.9, reached through vector…KEVEPSS 92%analysed8.8CVE-2020-16009Google Chrome V8 type confusion leads to heap corruptionGoogle Chrome before 86.0.4240.183 contains an inappropriate implementation in the V8 JavaScript engine that allows heap corruption via a crafted HTM…KEVEPSS 48%analysed8.8CVE-2019-13720Google Chrome WebAudio use-after-free allows heap corruptionGoogle Chrome before 78.0.3904.87 contains a use-after-free in WebAudio that a remote attacker can trigger with a crafted HTML page, leading to heap …KEVEPSS 49%analysed8.8CVE-2016-1646Google Chrome V8 Array.concat out-of-bounds readThe Array.prototype.concat implementation in Google V8 fails to properly consider element data types, causing an out-of-bounds read. It affects Googl…KEVEPSS 48%analysed

Source: NIST National Vulnerability Database (record CVE-2020-15211), CISA KEV, FIRST EPSS (scores of 2026-11-18). This page is refreshed as NVD updates the record.