Headline
Warning: PyTorch Models Vulnerable to Remote Code Execution via ShellTorch
Cybersecurity researchers have disclosed multiple critical security flaws in the TorchServe tool for serving and scaling PyTorch models that could be chained to achieve remote code execution on affected systems. Israel-based runtime application security company Oligo, which made the discovery, has coined the vulnerabilities ShellTorch. "These vulnerabilities […] can lead to a full chain Remote
Artificial Intelligence / Cyber Threat
Cybersecurity researchers have disclosed multiple critical security flaws in the TorchServe tool for serving and scaling PyTorch models that could be chained to achieve remote code execution on affected systems.
Israel-based runtime application security company Oligo, which made the discovery, has coined the vulnerabilities ShellTorch.
“These vulnerabilities […] can lead to a full chain Remote Code Execution (RCE), leaving countless thousands of services and end-users — including some of the world’s largest companies — open to unauthorized access and insertion of malicious AI models, and potentially a full server takeover,” security researchers Idan Levcovich, Guy Kaplan, and Gal Elbaz said.
The list of flaws, which have been addressed in version 0.8.2, is as follows -
- No CVE - Unauthenticated Management Interface API Misconfiguration (0.0.0.0)
- CVE-2023-43654 (CVSS score: 7.2) - A remote server-side request forgery (SSRF) that leads to remote code execution.
- CVE-2022-1471 (CVSS score: 9.9) - Use of an insecure version of the SnakeYAML open-source library that allows for unsafe deserialization of Java objects
Successful exploitation of the aforementioned flaws could allow an attacker to send a request to upload a malicious model from an actor-controlled address, leading to arbitrary code execution.
Put in other words, an attacker who can remotely access the management server can also upload a malicious model, which enables code execution without requiring any authentication on any default TorchServe server.
Even more troublingly, the shortcomings could be chained with CVE-2022-1471 to pave the way for code execution and full takeover of exposed instances.
“AI models can include a YAML file to declare their desired configuration, so by uploading a model with a maliciously crafted YAML file, we were able to trigger an unsafe deserialization attack that resulted in code execution on the machine,” the researchers said.
The severity of the issues has prompted Amazon Web Services (AWS) to issue an advisory urging customers using PyTorch inference Deep Learning Containers (DLC) 1.13.1, 2.0.0, or 2.0.1 in EC2, EKS, or ECS released prior to September 11, 2023, update to TorchServe version 0.8.2.
“Using the privileges granted by these vulnerabilities, it is possible to view, modify, steal, and delete AI models and sensitive data flowing into and from the target TorchServe server,” the researchers said.
“Making these vulnerabilities even more dangerous: when an attacker exploits the model serving server, they can access and alter sensitive data flowing in and out from the target TorchServe server, harming the trust and credibility of the application.”
Found this article interesting? Follow us on Twitter and LinkedIn to read more exclusive content we post.
Related news
Red Hat Security Advisory 2024-1353-03 - An update is now available for Red Hat Process Automation Manager. Issues addressed include code execution, denial of service, and deserialization vulnerabilities.
Atlassian has released software fixes to address four critical flaws in its software that, if successfully exploited, could result in remote code execution. The list of vulnerabilities is below - CVE-2022-1471 (CVSS score: 9.8) - Deserialization vulnerability in SnakeYAML library that can lead to remote code execution in multiple products CVE-2023-22522 (CVSS score
The PyTorch model server contains multiple vulnerabilities that can be chained together to permit an unauthenticated remote attacker arbitrary Java code execution. The first vulnerability is that the management interface is bound to all IP addresses and not just the loop back interface as the documentation suggests. The second vulnerability (CVE-2023-43654) allows attackers with access to the management interface to register MAR model files from arbitrary servers. The third vulnerability is that when an MAR file is loaded, it can contain a YAML configuration file that when deserialized by snakeyaml, can lead to loading an arbitrary Java class.
## Impact **Remote Server-Side Request Forgery (SSRF)** **Issue**: TorchServe default configuration lacks proper input validation, enabling third parties to invoke remote HTTP download requests and write files to the disk. This issue could be taken advantage of to compromise the integrity of the system and sensitive data. This issue is present in versions `0.1.0` to `0.8.1`. **Mitigation**: The user is able to load the model of their choice from any URL that they would like to use. The user of TorchServe is responsible for configuring both the [allowed_urls](https://github.com/pytorch/serve/blob/b3eced56b4d9d5d3b8597aa506a0bcf954d291bc/docs/configuration.md?plain=1#L296) and specifying the model URL to be used. A pull request to warn the user when the default value for `allowed_urls` is used has been merged - https://github.com/pytorch/serve/pull/2534. TorchServe release `0.8.2` includes this change. ## Patches ## TorchServe release 0.8.2 includes fixes to address the previou...
TorchServe is a tool for serving and scaling PyTorch models in production. TorchServe default configuration lacks proper input validation, enabling third parties to invoke remote HTTP download requests and write files to the disk. This issue could be taken advantage of to compromise the integrity of the system and sensitive data. This issue is present in versions 0.1.0 to 0.8.1. A user is able to load the model of their choice from any URL that they would like to use. The user of TorchServe is responsible for configuring both the allowed_urls and specifying the model URL to be used. A pull request to warn the user when the default value for allowed_urls is used has been merged in PR #2534. TorchServe release 0.8.2 includes this change. Users are advised to upgrade. There are no known workarounds for this issue.
A security update is now available for Red Hat JBoss Enterprise Application Platform 7.4. Red Hat Product Security has rated this update as having a security impact of Important. A Common Vulnerability Scoring System (CVSS) base score, which gives a detailed severity rating, is available for each vulnerability from the CVE link(s) in the References section.This content is licensed under the Creative Commons Attribution 4.0 International License (https://creativecommons.org/licenses/by/4.0/). If you distribute this content, or a modified version of it, you must provide attribution to Red Hat Inc. and provide a link to the original. Related CVEs: * CVE-2022-1471: A flaw was found in the SnakeYaml package. This flaw allows an attacker to benefit from remote code execution by sending malicious YAML content and this content being deserialized by the constructor. Deserialization is unsafe and leads to Remote Code Execution (RCE). * CVE-2022-4492: A flaw was found in undertow. The undertow c...
An update is now available for Red Hat OpenShift GitOps 1.7. Red Hat Product Security has rated this update as having a security impact of Moderate. A Common Vulnerability Scoring System (CVSS) base score, which gives a detailed severity rating, is available for each vulnerability from the CVE link(s) in the References section.This content is licensed under the Creative Commons Attribution 4.0 International License (https://creativecommons.org/licenses/by/4.0/). If you distribute this content, or a modified version of it, you must provide attribution to Red Hat Inc. and provide a link to the original. Related CVEs: * CVE-2022-41354: An information disclosure flaw was found in Argo CD. This issue may allow unauthorized users to enumerate application names by inspecting API error messages and could use the discovered application names as the starting point of another attack. For example, the attacker might use their knowledge of an application name to convince an administrator to grant ...
An update is now available for Red Hat OpenShift GitOps 1.6. Red Hat Product Security has rated this update as having a security impact of Moderate. A Common Vulnerability Scoring System (CVSS) base score, which gives a detailed severity rating, is available for each vulnerability from the CVE link(s) in the References section.This content is licensed under the Creative Commons Attribution 4.0 International License (https://creativecommons.org/licenses/by/4.0/). If you distribute this content, or a modified version of it, you must provide attribution to Red Hat Inc. and provide a link to the original. Related CVEs: * CVE-2022-41354: An information disclosure flaw was found in Argo CD. This issue may allow unauthorized users to enumerate application names by inspecting API error messages and could use the discovered application names as the starting point of another attack. For example, the attacker might use their knowledge of an application name to convince an administrator to grant ...
Red Hat Security Advisory 2023-1045-01 - Red Hat Single Sign-On 7.6 is a standalone server, based on the Keycloak project, that provides authentication and standards-based single sign-on capabilities for web and mobile applications. This release of Red Hat Single Sign-On 7.6.2 on RHEL 9 serves as a replacement for Red Hat Single Sign-On 7.6.1, and includes bug fixes and enhancements, which are documented in the Release Notes document linked to in the References. Issues addressed include code execution, cross site scripting, denial of service, deserialization, html injection, memory exhaustion, server-side request forgery, and traversal vulnerabilities.
Red Hat Security Advisory 2023-1043-01 - Red Hat Single Sign-On 7.6 is a standalone server, based on the Keycloak project, that provides authentication and standards-based single sign-on capabilities for web and mobile applications. This release of Red Hat Single Sign-On 7.6.2 on RHEL 7 serves as a replacement for Red Hat Single Sign-On 7.6.1, and includes bug fixes and enhancements, which are documented in the Release Notes document linked to in the References. Issues addressed include code execution, cross site scripting, denial of service, deserialization, html injection, memory exhaustion, server-side request forgery, and traversal vulnerabilities.
New Red Hat Single Sign-On 7.6.2 packages are now available for Red Hat Enterprise Linux 9. Red Hat Product Security has rated this update as having a security impact of Important. A Common Vulnerability Scoring System (CVSS) base score, which gives a detailed severity rating, is available for each vulnerability from the CVE link(s) in the References section.This content is licensed under the Creative Commons Attribution 4.0 International License (https://creativecommons.org/licenses/by/4.0/). If you distribute this content, or a modified version of it, you must provide attribution to Red Hat Inc. and provide a link to the original. Related CVEs: * CVE-2018-14040: In Bootstrap before 4.1.2, XSS is possible in the collapse data-parent attribute. * CVE-2018-14042: In Bootstrap before 4.1.2, XSS is possible in the data-container property of tooltip. * CVE-2019-11358: A Prototype Pollution vulnerability was found in jquery. Untrusted JSON passed to the `extend` function could lead to modi...
Red Hat OpenShift Container Platform release 4.10.52 is now available with updates to packages and images that fix several bugs and add enhancements. Red Hat Product Security has rated this update as having a security impact of Important. A Common Vulnerability Scoring System (CVSS) base score, which gives a detailed severity rating, is available for each vulnerability from the CVE link(s) in the References section.This content is licensed under the Creative Commons Attribution 4.0 International License (https://creativecommons.org/licenses/by/4.0/). If you distribute this content, or a modified version of it, you must provide attribution to Red Hat Inc. and provide a link to the original. Related CVEs: * CVE-2022-1471: A flaw was found in the SnakeYaml package. This flaw allows an attacker to benefit from remote code execution by sending malicious YAML content and this content being deserialized by the constructor. Deserialization is unsafe and leads to Remote Code Execution (RCE). *...
An update for prometheus-jmx-exporter is now available for Red Hat Enterprise Linux 8. Red Hat Product Security has rated this update as having a security impact of Important. A Common Vulnerability Scoring System (CVSS) base score, which gives a detailed severity rating, is available for each vulnerability from the CVE link(s) in the References section.This content is licensed under the Creative Commons Attribution 4.0 International License (https://creativecommons.org/licenses/by/4.0/). If you distribute this content, or a modified version of it, you must provide attribution to Red Hat Inc. and provide a link to the original. Related CVEs: * CVE-2022-1471: SnakeYaml: Constructor Deserialization Remote Code Execution
### Summary SnakeYaml's `Constructor` class, which inherits from `SafeConstructor`, allows any type be deserialized given the following line: new Yaml(new Constructor(TestDataClass.class)).load(yamlContent); Types do not have to match the types of properties in the target class. A `ConstructorException` is thrown, but only after a malicious payload is deserialized. ### Severity High, lack of type checks during deserialization allows remote code execution. ### Proof of Concept Execute `bash run.sh`. The PoC uses Constructor to deserialize a payload for RCE. RCE is demonstrated by using a payload which performs a http request to http://127.0.0.1:8000. Example output of successful run of proof of concept: ``` $ bash run.sh [+] Downloading snakeyaml if needed [+] Starting mock HTTP server on 127.0.0.1:8000 to demonstrate RCE nc: no process found [+] Compiling and running Proof of Concept, which a payload that sends a HTTP request to mock web server. [+] An exception is expected. Exc...
SnakeYaml's Constructor() class does not restrict types which can be instantiated during deserialization. Deserializing yaml content provided by an attacker can lead to remote code execution. We recommend using SnakeYaml's SafeConsturctor when parsing untrusted content to restrict deserialization.