A command injection vulnerability exists in mlflow/mlflow versions before v3.7.0, specifically in the `mlflow/sagemaker/__init__.py` file at lines 161-167. The vulnerability arises from the direct interpolation of user-supplied container image names into shell commands without proper sanitization, which are then executed using `os.system()`. This allows attackers to execute arbitrary commands by supplying malicious input through the `--container` parameter of the CLI. The issue affects environments where MLflow is used, including development setups, CI/CD pipelines, and cloud deployments.

Insufficient validation of node IDs in Qt SVG module allows arbitrary QML/JavaScript code injection when loading malicious SVG files through the VectorImage component in Qt Quick. While QML execution is typically more restricted than native code execution, this could still lead to denial of service, information disclosure, or other impacts depending on the application's privilege level and data access.

A flaw was found in the GTK library. Under certain conditions, it is possible for a library to be injected into a GTK application from the current working directory.

foomaticrip.c in foomatic-rip in foomatic-filters in Foomatic 4.0.6 allows remote attackers to execute arbitrary code via a crafted *FoomaticRIPCommandLine field in a .ppd file, a different vulnerability than CVE-2011-2697.

vm2 is an open source vm/sandbox for Node.js. Prior to version 3.11.0, VM2 suffers from a sandbox breakout vulnerability through the inspect function. This allows attackers to write code which can escape from the VM2 sandbox and execute arbitrary commands on the host system. This issue has been patched in version 3.11.0.

A flaw was found in IPA, all 4.6.x versions before 4.6.7, all 4.7.x versions before 4.7.4 and all 4.8.x versions before 4.8.3, in the way the internal function ber_scanf() was used in some components of the IPA server, which parsed kerberos key data. An unauthenticated attacker who could trigger parsing of the krb principal key could cause the IPA server to crash or in some conditions, cause arbitrary code to be executed on the server hosting the IPA server.

A type confusion in the nas_message_decode function of Magma <= 1.8.0 (fixed in v1.9 commit 08472ba98b8321f802e95f5622fa90fec2dea486) allows attackers to execute arbitrary code or cause a Denial of Service (DoS) via a crafted NAS packet.

Vim is an open source, command line text editor. Prior to version 9.2.0597, Vim's Python omni-completion executes reconstructed function and class definitions from the current buffer with exec() as part of populating the completion dictionary. Python evaluates function default values, parameter annotations, and class base expressions at definition time, so a hostile buffer can execute attacker-controlled Python expressions during omni-completion. The existing g:pythoncomplete_allow_import mitigation (GHSA-52mc-rq6p-rc7c) does not cover this path, because the attacker-controlled code is not a harvested import/from statement. This issue has been patched in version 9.2.0597.

Impact: The fix for CVE-2021-23337 (https://github.com/advisories/GHSA-35jh-r3h4-6jhm) added validation for the variable option in _.template but did not apply the same validation to options.imports key names. Both paths flow into the same Function() constructor sink. When an application passes untrusted input as options.imports key names, an attacker can inject default-parameter expressions that execute arbitrary code at template compilation time. Additionally, _.template uses assignInWith to merge imports, which enumerates inherited properties via for..in. If Object.prototype has been polluted by any other vector, the polluted keys are copied into the imports object and passed to Function(). Patches: Users should upgrade to version 4.18.0. Workarounds: Do not pass untrusted input as key names in options.imports. Only use developer-controlled, static key names.

The Mock software contains a vulnerability wherein an attacker could potentially exploit privilege escalation, enabling the execution of arbitrary code with root user privileges. This weakness stems from the absence of proper sandboxing during the expansion and execution of Jinja2 templates, which may be included in certain configuration parameters. While the Mock documentation advises treating users added to the mock group as privileged, certain build systems invoking mock on behalf of users might inadvertently permit less privileged users to define configuration tags. These tags could then be passed as parameters to mock during execution, potentially leading to the utilization of Jinja2 templates for remote privilege escalation and the execution of arbitrary code as the root user on the build server.

IO::Compress versions before 2.220 for Perl can execute arbitrary code in File::GlobMapper via an attacker-controlled output glob. _parseOutputGlob() wraps the caller-supplied output glob string in double quotes and stores it in the parser state; _getFiles() then runs the stored expression through eval STRING. A literal double quote in the output glob closes the dquote wrapper, and the characters that follow are evaluated as Perl. Arbitrary Perl in the output glob executes at the calling process's privilege.

Vim is an open source, command line text editor. Prior to version 9.2.0495, a Vimscript code injection vulnerability exists in s:NetrwBookHistSave() in the netrw plugin (runtime/pack/dist/opt/netrw/autoload/netrw.vim) when serializing browsed directory paths to the history file ~/.vim/.netrwhist. A directory name derived from the filesystem is interpolated into a single-quoted Vimscript string literal without escaping embedded single quotes, allowing a crafted directory name to break out of the string context and execute arbitrary Vimscript, including shell commands via system() and :!, the next time the history file is sourced. This issue has been patched in version 9.2.0495.

A pre-authentication, code injection vulnerability in version 1.0.0 or later of the ChromaDB Python project allows an unauthenticated attacker to run arbitrary code on the server by sending a malicious model repository and trust_remote_code set to true in the /api/v2/tenants/{tenant}/databases/{db}/collections endpoint.

vm2 is an open source vm/sandbox for Node.js. Prior to 3.11.0, it is possible to obtain the host Object. There are various ways to use the host Object, to escape the sandbox, one example would be using HostObject.getOwnPropertySymbols to obtain Symbol(nodejs.util.inspect.custom). This vulnerability is fixed in 3.11.0.

vm2 is an open source vm/sandbox for Node.js. From 3.9.6 to 3.10.5, vm2's bridge exposes mutable proxies for real host-realm intrinsic prototypes and then forwards sandbox writes into the underlying host objects with otherReflectSet() and otherReflectDefineProperty(), which lets attacker-controlled JavaScript running in a default VM or inherited NodeVM mutate shared host Object.prototype, Array.prototype, and Function.prototype from inside the sandbox This vulnerability is fixed in 3.11.0.

vm2 is an open source vm/sandbox for Node.js. Prior to 3.11.0, It is possible to reach BaseHandler.getPrototypeOf, which can be used to get arbitrary prototypes. This vulnerability is fixed in 3.11.0.

Diffusers is the a library for pretrained diffusion models. Prior to 0.38.0, a trust_remote_code bypass in DiffusionPipeline.from_pretrained allows arbitrary remote code execution despite the user passing trust_remote_code=False (or omitting it, which is the default). The vulnerability has three variants, all sharing the same root cause — the trust_remote_code gate was implemented inside DiffusionPipeline.download() rather than at the actual dynamic-module load site, so any code path that bypassed or short-circuited download() also bypassed the security check. DiffusionPipeline.from_pretrained('repoA', custom_pipeline='attacker/repoB', trust_remote_code=False) — the gate evaluated against repoA's file list rather than repoB's, so repoB's pipeline.py was loaded and executed. DiffusionPipeline.from_pretrained('/local/snapshot', custom_pipeline='attacker/repoB', trust_remote_code=False) — the local-path branch never invoked download(), so the gate was never reached and remote code from repoB executed. DiffusionPipeline.from_pretrained('/local/snapshot', trust_remote_code=False) where the snapshot contains custom component files (e.g. unet/my_unet_model.py) referenced from model_index.json — same root cause; the local path skipped download() and custom component code executed. This vulnerability is fixed in 0.38.0.

Axios is a promise based HTTP client for the browser and Node.js. From 0.19.0 to before 0.31.1 and 1.15.2, Axios contains prototype-pollution gadgets in request config processing. If another vulnerability in the same JavaScript process has already polluted Object.prototype.transformResponse, affected Axios versions may treat that inherited value as request configuration or as an option validator. Axios does not itself create the prototype pollution. Exploitability requires a separate prototype-pollution vulnerability or equivalent attacker control over Object.prototype before Axios creates a request. This vulnerability is fixed in 0.31.1 and 1.15.2.

Docling simplifies document processing by parsing diverse formats and providing integrations with the generative AI ecosystem. FIn versions >= 2.82.0, < 2.91.0, if the HTML backend was explicitly configured for rendering (rendering option by default deactivated), then the Playwright-based rendering feature could allow JavaScript execution and unrestricted network access when processing untrusted HTML documents. An attacker could craft malicious HTML that executes arbitrary JavaScript in the rendering context or makes unauthorized network requests to internal services, potentially leading to SSRF attacks, data exfiltration, or remote code execution in the rendering environment. This vulnerability is fixed in 2.91.0.

Backstage is an open platform for building developer portals. The Backstage scaffolder-backend plugin uses a templating library that requires sandbox, as it by design allows for code injection. The library used for this sandbox so far has been `vm2`, but in light of several past vulnerabilities and existing vulnerabilities that may not have a fix, the plugin has switched to using a different sandbox library. A malicious actor with write access to a registered scaffolder template could manipulate the template in a way that allows for remote code execution on the scaffolder-backend instance. This was only exploitable in the template YAML definition itself and not by user input data. This is vulnerability is fixed in version 1.15.0 of `@backstage/plugin-scaffolder-backend`.

Rclone is a command-line program to sync files and directories to and from different cloud storage providers. Starting in version 1.48.0 and prior to version 1.73.5, the RC endpoint `operations/fsinfo` is exposed without `AuthRequired: true` and accepts attacker-controlled `fs` input. Because `rc.GetFs(...)` supports inline backend definitions, an unauthenticated attacker can instantiate an attacker-controlled backend on demand. For the WebDAV backend, `bearer_token_command` is executed during backend initialization, making single-request unauthenticated local command execution possible on reachable RC deployments without global HTTP authentication. Version 1.73.5 patches the issue.

Contour is a Kubernetes ingress controller using Envoy proxy. From v1.19.0 to before v1.33.4, v1.32.5, and v1.31.6, Contour's Cookie Rewriting feature is vulnerable to Lua code injection. An attacker with RBAC permissions to create or modify HTTPProxy resources can craft a malicious value in spec.routes[].cookieRewritePolicies[].pathRewrite.value or spec.routes[].services[].cookieRewritePolicies[].pathRewrite.value that results in arbitrary code execution in the Envoy proxy. The cookie rewriting feature is internally implemented using Envoy's HTTP Lua filter. User-controlled values are interpolated into Lua source code using Go text/template without sufficient sanitization. The injected code only executes when processing traffic on the attacker's own route, which they already control. However, since Envoy runs as shared infrastructure, the injected code can also read Envoy's xDS client credentials from the filesystem or cause denial of service for other tenants sharing the Envoy instance. This vulnerability is fixed in v1.33.4, v1.32.5, and v1.31.6.

vLLM is an inference and serving engine for large language models (LLMs). Prior to 0.22.0, an assert-based security check in vLLM's activation function loading allows any unauthenticated attacker to achieve arbitrary code execution on the server by publishing a malicious HuggingFace model, when vLLM runs in Python optimized mode (python -O or PYTHONOPTIMIZE=1). This vulnerability is fixed in 0.22.0.

A flaw was found in libinput. A local attacker who can place a specially crafted Lua bytecode file in certain system or user configuration directories can bypass security restrictions. This allows the attacker to run unauthorized code with the same permissions as the program using libinput, such as a graphical compositor. This could lead to the attacker monitoring keyboard input and sending that information to an external location.

Kedro is a toolbox for production-ready data science. Prior to 1.3.0, Kedro allows the logging configuration file path to be set via the KEDRO_LOGGING_CONFIG environment variable and loads it without validation. The logging configuration schema supports the special () key, which enables arbitrary callable instantiation. An attacker can exploit this to execute arbitrary system commands during application startup. This is a critical remote code execution (RCE) vulnerability caused by unsafe use of logging.config.dictConfig() with user-controlled input. This vulnerability is fixed in 1.3.0.

Handlebars provides the power necessary to let users build semantic templates. In versions 4.0.0 through 4.7.8, `Handlebars.compile()` accepts a pre-parsed AST object in addition to a template string. The `value` field of a `NumberLiteral` AST node is emitted directly into the generated JavaScript without quoting or sanitization. An attacker who can supply a crafted AST to `compile()` can therefore inject and execute arbitrary JavaScript, leading to Remote Code Execution on the server. Version 4.7.9 fixes the issue. Some workarounds are available. Validate input type before calling `Handlebars.compile()`; ensure the argument is always a `string`, never a plain object or JSON-deserialized value. Use the Handlebars runtime-only build (`handlebars/runtime`) on the server if templates are pre-compiled at build time; `compile()` will be unavailable.

Handlebars provides the power necessary to let users build semantic templates. In versions 4.0.0 through 4.7.8, the `@partial-block` special variable is stored in the template data context and is reachable and mutable from within a template via helpers that accept arbitrary objects. When a helper overwrites `@partial-block` with a crafted Handlebars AST, a subsequent invocation of `{{> @partial-block}}` compiles and executes that AST, enabling arbitrary JavaScript execution on the server. Version 4.7.9 fixes the issue. Some workarounds are available. First, use the runtime-only build (`require('handlebars/runtime')`). The `compile()` method is absent, eliminating the vulnerable fallback path. Second, audit registered helpers for any that write arbitrary values to context objects. Helpers should treat context data as read-only. Third, avoid registering helpers from third-party packages (such as `handlebars-helpers`) in contexts where templates or context data can be influenced by untrusted input.

Handlebars provides the power necessary to let users build semantic templates. In versions 4.0.0 through 4.7.8, a crafted object placed in the template context can bypass all conditional guards in `resolvePartial()` and cause `invokePartial()` to return `undefined`. The Handlebars runtime then treats the unresolved partial as a source that needs to be compiled, passing the crafted object to `env.compile()`. Because the object is a valid Handlebars AST containing injected code, the generated JavaScript executes arbitrary commands on the server. The attack requires the adversary to control a value that can be returned by a dynamic partial lookup. Version 4.7.9 fixes the issue. Some workarounds are available. First, use the runtime-only build (`require('handlebars/runtime')`). Without `compile()`, the fallback compilation path in `invokePartial` is unreachable. Second, sanitize context data before rendering: Ensure no value in the context is a non-primitive object that could be passed to a dynamic partial. Third, avoid dynamic partial lookups (`{{> (lookup ...)}}`) when context data is user-controlled.

Handlebars provides the power necessary to let users build semantic templates. In versions 4.0.0 through 4.7.8, the Handlebars CLI precompiler (`bin/handlebars` / `lib/precompiler.js`) concatenates user-controlled strings — template file names and several CLI options — directly into the JavaScript it emits, without any escaping or sanitization. An attacker who can influence template filenames or CLI arguments can inject arbitrary JavaScript that executes when the generated bundle is loaded in Node.js or a browser. Version 4.7.9 fixes the issue. Some workarounds are available. First, validate all CLI inputs before invoking the precompiler. Reject filenames and option values that contain characters with JavaScript string-escaping significance (`"`, `'`, `;`, etc.). Second, use a fixed, trusted namespace string passed via a configuration file rather than command-line arguments in automated pipelines. Third, run the precompiler in a sandboxed environment (container with no write access to sensitive paths) to limit the impact of successful exploitation. Fourth, audit template filenames in any repository or package that is consumed by an automated build pipeline.

Happy DOM is a JavaScript implementation of a web browser without its graphical user interface. In versions 15.10.0 through 20.8.7, a code injection vulnerability in `ECMAScriptModuleCompiler` allows an attacker to achieve Remote Code Execution (RCE) by injecting arbitrary JavaScript expressions inside `export { }` declarations in ES module scripts processed by happy-dom. The compiler directly interpolates unsanitized content into generated code as an executable expression, and the quote filter does not strip backticks, allowing template literal-based payloads to bypass sanitization. Version 20.8.8 fixes the issue.

SimpleEval is a library for adding evaluatable expressions into python projects. Prior to 1.0.5, objects (including modules) can leak dangerous modules through to direct access inside the sandbox. If the objects you've passed in as names to SimpleEval have modules or other disallowed / dangerous objects available as attrs. Additionally, dangerous functions or modules could be accessed by passing them as callbacks to other safe functions to call. The latest version 1.0.5 has this issue fixed. This vulnerability is fixed in 1.0.5.

python jsonpickle 2.0.0 contains a remote code execution vulnerability that allows attackers to execute arbitrary Python commands by deserializing malicious JSON payloads containing py/repr objects. Attackers can craft JSON strings with py/repr directives that invoke the eval function during deserialization to execute arbitrary code.

Podman is a tool for managing OCI containers and pods. Versions 4.8.0 through 5.8.1 contain a command injection vulnerability in the HyperV machine backend in pkg/machine/hyperv/stubber.go, where the VM image path is inserted into a PowerShell double-quoted string without sanitization, allowing $() subexpression injection. Because PowerShell evaluates subexpressions inside double-quoted strings before executing the outer command, an attacker who can control the VM image path through a crafted machine name or image directory can execute arbitrary PowerShell commands with the privileges of the Podman process. On typical Windows installations this means SYSTEM-level code execution, and only Windows is affected as the code is exclusive to the HyperV backend. This issue has been patched in version 5.8.2.

Spinnaker is an open source, multi-cloud continuous delivery platform. Echo like some other services, uses SPeL (Spring Expression Language) to process information - specifically around expected artifacts. In versions prior to 2026.1.0, 2026.0.1, 2025.4.2, and 2025.3.2, unlike orca, it was NOT restricting that context to a set of trusted classes, but allowing FULL JVM access. This enabled a user to use arbitrary java classes which allow deep access to the system. This enabled the ability to invoke commands, access files, etc. Versions 2026.1.0, 2026.0.1, 2025.4.2, and 2025.3.2 contain a patch. As a workaround, disable echo entirely.

Locutus brings stdlibs of other programming languages to JavaScript for educational purposes. Prior to 3.0.14, the create_function(args, code) function passes both parameters directly to the Function constructor without any sanitization, allowing arbitrary code execution. This is distinct from CVE-2026-29091 which was call_user_func_array using eval() in v2.x. This finding affects create_function using new Function() in v3.x. This vulnerability is fixed in 3.0.14.

jsPDF is a library to generate PDFs in JavaScript. Prior to version 4.2.1, user control of arguments of the `createAnnotation` method allows users to inject arbitrary PDF objects, such as JavaScript actions. If given the possibility to pass unsanitized input to the following method, a user can inject arbitrary PDF objects, such as JavaScript actions, which might trigger when the PDF is opened or interacted with the `createAnnotation`: `color` parameter. The vulnerability has been fixed in jsPDF@4.2.1. As a workaround, sanitize user input before passing it to the vulnerable API members.

Locutus brings stdlibs of other programming languages to JavaScript for educational purposes. Prior to version 3.0.0, a remote code execution (RCE) flaw was discovered in the locutus project, specifically within the call_user_func_array function implementation. The vulnerability allows an attacker to inject arbitrary JavaScript code into the application's runtime environment. This issue stems from an insecure implementation of the call_user_func_array function (and its wrapper call_user_func), which fails to properly validate all components of a callback array before passing them to eval(). This issue has been patched in version 3.0.0.

The llm CLI tool thru 0.27.1 contains a critical code injection vulnerability via its --functions command-line argument. This argument is intended to allow users to provide custom Python function definitions. However, the tool directly executes the provided code using the unsafe exec() function without any sanitization, sandboxing, or security restrictions. An attacker can exploit this by crafting a malicious llm command with arbitrary Python code in the --functions argument and using social engineering to trick a victim into running it. This leads to arbitrary code execution on the victim's system, potentially granting the attacker full control.

c3p0, a JDBC Connection pooling library, is vulnerable to attack via maliciously crafted Java-serialized objects and `javax.naming.Reference` instances. Several c3p0 `ConnectionPoolDataSource` implementations have a property called `userOverridesAsString` which conceptually represents a `Map<String,Map<String,String>>`. Prior to v0.12.0, that property was maintained as a hex-encoded serialized object. Any attacker able to reset this property, on an existing `ConnectionPoolDataSource` or via maliciously crafted serialized objects or `javax.naming.Reference` instances could be tailored execute unexpected code on the application's `CLASSPATH`. The danger of this vulnerability was strongly magnified by vulnerabilities in c3p0's main dependency, mchange-commons-java. This library includes code that mirrors early implementations of JNDI functionality, including ungated support for remote `factoryClassLocation` values. Attackers could set c3p0's `userOverridesAsString` hex-encoded serialized objects that include objects "indirectly serialized" via JNDI references. Deserialization of those objects and dereferencing of the embedded `javax.naming.Reference` objects could provoke download and execution of malicious code from a remote `factoryClassLocation`. Although hazard presented by c3p0's vulnerabilites are exarcerbated by vulnerabilities in mchange-commons-java, use of Java-serialized-object hex as the format for a writable Java-Bean property, of objects that may be exposed across JNDI interfaces, represents a serious independent fragility. The `userOverridesAsString` property of c3p0 `ConnectionPoolDataSource` classes has been reimplemented to use a safe CSV-based format, rather than rely upon potentially dangerous Java object deserialization. c3p0-0.12.0+ and above depend upon mchange-commons-java 0.4.0+, which gates support for remote `factoryClassLocation` values by configuration parameters that default to restrictive values. c3p0 additionally enforces the new mchange-commons-java `com.mchange.v2.naming.nameGuardClassName` to prevent injection of unexpected, potentially remote JNDI names. There is no supported workaround for versions of c3p0 prior to 0.12.0.

A flaw was found in the libreswan client plugin for NetworkManager (NetkworkManager-libreswan), where it fails to properly sanitize the VPN configuration from the local unprivileged user. In this configuration, composed by a key-value format, the plugin fails to escape special characters, leading the application to interpret values as keys. One of the most critical parameters that could be abused by a malicious user is the `leftupdown`key. This key takes an executable command as a value and is used to specify what executes as a callback in NetworkManager-libreswan to retrieve configuration settings back to NetworkManager. As NetworkManager uses Polkit to allow an unprivileged user to control the system's network configuration, a malicious actor could achieve local privilege escalation and potential code execution as root in the targeted machine by creating a malicious configuration.

A chained attack via SQL Expressions and a Grafana Enterprise plugin can lead to a remote arbitrary code execution impact (RCE). This is enabled by a feature in Grafana (OSS), so all users are always recommended to update to avoid future attack vectors going this path. Only instances with the sqlExpressions feature toggle enabled are vulnerable. Only instances in the following version ranges are affected: - 11.6.0 (inclusive) to 11.6.14 (exclusive): 11.6.14 has the fix. 11.5 and below are not affected. - 12.0.0 (inclusive) to 12.1.10 (exclusive): 12.1.10 has the fix. 12.0 did not receive an update, as it is end-of-life. - 12.2.0 (inclusive) to 12.2.8 (exclusive): 12.2.8 has the fix. - 12.3.0 (inclusive) to 12.3.6 (exclusive): 12.3.6 has the fix. - 12.4.0 (inclusive) to 12.4.2 (exclusive): 12.4.2 has the fix. 13.0.0 and above also have the fix: no v13 release is affected.

vm2 is an open source vm/sandbox for Node.js. Prior to version 3.11.0, SuppressedError allows attackers to escape the sandbox and run arbitrary code. This issue has been patched in version 3.11.0.

jsPDF is a library to generate PDFs in JavaScript. Prior to 4.2.0, user control of the argument of the `addJS` method allows an attacker to inject arbitrary PDF objects into the generated document. By crafting a payload that escapes the JavaScript string delimiter, an attacker can execute malicious actions or alter the document structure, impacting any user who opens the generated PDF. The vulnerability has been fixed in jspdf@4.2.0. As a workaround, escape parentheses in user-provided JavaScript code before passing them to the `addJS` method.

PyTorch is a Python package that provides tensor computation. Prior to version 2.10.0, a vulnerability in PyTorch's `weights_only` unpickler allows an attacker to craft a malicious checkpoint file (`.pth`) that, when loaded with `torch.load(..., weights_only=True)`, can corrupt memory and potentially lead to arbitrary code execution. Version 2.10.0 fixes the issue.

vm2 is an open source vm/sandbox for Node.js. Prior to version 3.11.0, VM2 suffers from a sandbox breakout vulnerability. This allows attackers to write code which can escape from the VM2 sandbox and execute arbitrary commands on the host system. This issue has been patched in version 3.11.0.