vLLM, an inference and serving engine for large language models (LLMs), has an issue in versions 0.6.5 through 0.8.4 that ONLY impacts environments using the `PyNcclPipe` KV cache transfer integration with the V0 engine. No other configurations are affected. vLLM supports the use of the `PyNcclPipe` class to establish a peer-to-peer communication domain for data transmission between distributed nodes. The GPU-side KV-Cache transmission is implemented through the `PyNcclCommunicator` class, while CPU-side control message passing is handled via the `send_obj` and `recv_obj` methods on the CPU side.​ The intention was that this interface should only be exposed to a private network using the IP address specified by the `--kv-ip` CLI parameter. The vLLM documentation covers how this must be limited to a secured network. The default and intentional behavior from PyTorch is that the `TCPStore` interface listens on ALL interfaces, regardless of what IP address is provided. The IP address given was only used as a client-side address to use. vLLM was fixed to use a workaround to force the `TCPStore` instance to bind its socket to a specified private interface. As of version 0.8.5, vLLM limits the `TCPStore` socket to the private interface as configured.

vllm-project vllm version 0.6.0 contains a vulnerability in the AsyncEngineRPCServer() RPC server entrypoints. The core functionality run_server_loop() calls the function _make_handler_coro(), which directly uses cloudpickle.loads() on received messages without any sanitization. This can result in remote code execution by deserializing malicious pickle data.

vLLM is a high-throughput and memory-efficient inference and serving engine for LLMs. When vLLM is configured to use Mooncake, unsafe deserialization exposed directly over ZMQ/TCP on all network interfaces will allow attackers to execute remote code on distributed hosts. This is a remote code execution vulnerability impacting any deployments using Mooncake to distribute KV across distributed hosts. This vulnerability is fixed in 0.8.0.

vLLM is an inference and serving engine for large language models. In a multi-node vLLM deployment using the V0 engine, vLLM uses ZeroMQ for some multi-node communication purposes. The secondary vLLM hosts open a `SUB` ZeroMQ socket and connect to an `XPUB` socket on the primary vLLM host. When data is received on this `SUB` socket, it is deserialized with `pickle`. This is unsafe, as it can be abused to execute code on a remote machine. Since the vulnerability exists in a client that connects to the primary vLLM host, this vulnerability serves as an escalation point. If the primary vLLM host is compromised, this vulnerability could be used to compromise the rest of the hosts in the vLLM deployment. Attackers could also use other means to exploit the vulnerability without requiring access to the primary vLLM host. One example would be the use of ARP cache poisoning to redirect traffic to a malicious endpoint used to deliver a payload with arbitrary code to execute on the target machine. Note that this issue only affects the V0 engine, which has been off by default since v0.8.0. Further, the issue only applies to a deployment using tensor parallelism across multiple hosts, which we do not expect to be a common deployment pattern. Since V0 is has been off by default since v0.8.0 and the fix is fairly invasive, the maintainers of vLLM have decided not to fix this issue. Instead, the maintainers recommend that users ensure their environment is on a secure network in case this pattern is in use. The V1 engine is not affected by this issue.

vLLM is a library for LLM inference and serving. vllm/model_executor/weight_utils.py implements hf_model_weights_iterator to load the model checkpoint, which is downloaded from huggingface. It uses the torch.load function and the weights_only parameter defaults to False. When torch.load loads malicious pickle data, it will execute arbitrary code during unpickling. This vulnerability is fixed in v0.7.0.

vllm-project vllm version v0.6.2 contains a vulnerability in the MessageQueue.dequeue() API function. The function uses pickle.loads to parse received sockets directly, leading to a remote code execution vulnerability. An attacker can exploit this by sending a malicious payload to the MessageQueue, causing the victim's machine to execute arbitrary code.

dotPDN Paint.NET before 4.1.2 allows Deserialization of Untrusted Data (issue 1 of 2).

​Delta Electronics InfraSuite Device Master versions prior to 1.0.7 contains classes that cannot be deserialized, which could allow an attack to remotely execute arbitrary code.

A YAML deserialization vulnerability was found in the Robot Operating System (ROS) 'dynparam', a command-line tool for getting, setting, and deleting parameters of a dynamically configurable node, affecting ROS distributions Noetic and earlier. The issue is caused by the use of the yaml.load() function in the 'set' and 'get' verbs, and allows for the creation of arbitrary Python objects. Through this flaw, a local or remote user can craft and execute arbitrary Python code.

The Feed Them Social – for Twitter feed, Youtube and more plugin for WordPress is vulnerable to deserialization of untrusted input via the 'fts_url' parameter in versions up to, and including 2.9.8.5. This makes it possible for unauthenticated attackers to call files using a PHAR wrapper that will deserialize the data and call arbitrary PHP Objects that can be used to perform a variety of malicious actions granted a POP chain is also present. It also requires that the attacker is successful in uploading a file with the serialized payload.

xxl-rpc v1.7.0 was discovered to contain a deserialization vulnerability via the component com.xxl.rpc.core.remoting.net.impl.netty.codec.NettyDecode#decode.

A vulnerability was found in Zhong Bang CRMEB up to 4.6.0. It has been declared as problematic. Affected by this vulnerability is the function put_image of the file api/controller/v1/PublicController.php. The manipulation leads to deserialization. The attack can be launched remotely. The exploit has been disclosed to the public and may be used. The identifier VDB-231505 was assigned to this vulnerability. NOTE: The vendor was contacted early about this disclosure but did not respond in any way.

IBM InfoSphere Information Server 11.7 is affected by a remote code execution vulnerability due to insecure deserialization in an RMI service. IBM X-Force ID: 255285.

The Ultimate Store Kit Elementor Addons, Woocommerce Builder, EDD Builder, Elementor Store Builder, Product Grid, Product Table, Woocommerce Slider plugin is vulnerable to PHP Object Injection via deserialization of untrusted input via the _ultimate_store_kit_compare_products cookie in versions up to , and including, 1.6.4. This makes it possible for an unauthenticated attacker to inject a PHP Object. No POP chain is present in the vulnerable plugin. If a POP chain is present via an additional plugin or theme installed on the target system, it could allow the attacker or above to delete arbitrary files, retrieve sensitive data, or execute code.

The Dataprobe iBoot PDU running firmware version 1.43.03312023 or earlier is vulnerable to authentication bypass. By manipulating the IP address field in the "iBootPduSiteAuth" cookie, a malicious agent can direct the device to connect to a rouge database.Successful exploitation allows the malicious agent to take actions with administrator privileges including, but not limited to, manipulating power levels, modifying user accounts, and exporting confidential user information

Deserialization of Untrusted Data vulnerability in NIX Solutions Ltd NIX Anti-Spam Light allows Object Injection.This issue affects NIX Anti-Spam Light: from n/a through 0.0.4.

Deserialization of untrusted data in IPC and Parquet readers in the Apache Arrow R package versions 4.0.0 through 16.1.0 allows arbitrary code execution. An application is vulnerable if it reads Arrow IPC, Feather or Parquet data from untrusted sources (for example, user-supplied input files). This vulnerability only affects the arrow R package, not other Apache Arrow implementations or bindings unless those bindings are specifically used via the R package (for example, an R application that embeds a Python interpreter and uses PyArrow to read files from untrusted sources is still vulnerable if the arrow R package is an affected version). It is recommended that users of the arrow R package upgrade to 17.0.0 or later. Similarly, it is recommended that downstream libraries upgrade their dependency requirements to arrow 17.0.0 or later. If using an affected version of the package, untrusted data can read into a Table and its internal to_data_frame() method can be used as a workaround (e.g., read_parquet(..., as_data_frame = FALSE)$to_data_frame()). This issue affects the Apache Arrow R package: from 4.0.0 through 16.1.0. Users are recommended to upgrade to version 17.0.0, which fixes the issue.

The ZoomSounds - WordPress Wave Audio Player with Playlist plugin for WordPress is vulnerable to PHP Object Injection in all versions up to, and including, 6.91 via deserialization of untrusted input from the 'margs' parameter. This makes it possible for unauthenticated attackers to inject a PHP Object. No known POP chain is present in the vulnerable software, which means this vulnerability has no impact unless another plugin or theme containing a POP chain is installed on the site. If a POP chain is present via an additional plugin or theme installed on the target system, it may allow the attacker to perform actions like delete arbitrary files, retrieve sensitive data, or execute code depending on the POP chain present.

A vulnerability was found in Zhong Bang CRMEB up to 4.6.0 and classified as critical. This issue affects some unknown processing of the file /api/wechat/app_auth of the component Image Upload. The manipulation leads to deserialization. The exploit has been disclosed to the public and may be used. The associated identifier of this vulnerability is VDB-231503. NOTE: The vendor was contacted early about this disclosure but did not respond in any way.

Deserialization of Untrusted Data vulnerability in Phoenixheart Referrer Detector allows Object Injection.This issue affects Referrer Detector: from n/a through 4.2.1.0.

Deserialization of Untrusted Data vulnerability in GiveWP GiveWP – Donation Plugin and Fundraising Platform.This issue affects GiveWP – Donation Plugin and Fundraising Platform: from n/a through 2.25.3.

An XML Deserialization vulnerability in glazedlists v1.11.0 allows an attacker to execute arbitrary code via the BeanXMLByteCoder.decode() parameter.

The s2Member Pro plugin for WordPress is vulnerable to PHP Object Injection in all versions up to, and including, 241216 via deserialization of untrusted input from the 's2member_pro_remote_op' vulnerable parameter. This makes it possible for unauthenticated attackers to inject a PHP Object. No known POP chain is present in the vulnerable software. If a POP chain is present via an additional plugin or theme installed on the target system, it could allow the attacker to delete arbitrary files, retrieve sensitive data, or execute code.

Deserialization of Untrusted Data vulnerability in WP MEDIA SAS Search & Replace search-and-replace.This issue affects Search & Replace: from n/a through 3.2.2.

Deserialization of untrusted data in Microsoft Messaging Queuing Service in Medtronic's Paceart Optima versions 1.11 and earlier on Windows allows an unauthorized user to impact a healthcare delivery organization’s Paceart Optima system cardiac device causing data to be deleted, stolen, or modified, or the Paceart Optima system being used for further network penetration via network connectivity.

A deserialization vulnerability existed when decode a malicious package.This issue affects Apache Dubbo: from 3.1.0 through 3.1.10, from 3.2.0 through 3.2.4. Users are recommended to upgrade to the latest version, which fixes the issue.

In Apache Linkis <=1.3.1, due to the lack of effective filtering of parameters, an attacker configuring malicious Mysql JDBC parameters in JDBC EengineConn Module will trigger a deserialization vulnerability and eventually lead to remote code execution. Therefore, the parameters in the Mysql JDBC URL should be blacklisted. Versions of Apache Linkis <= 1.3.0 will be affected. We recommend users upgrade the version of Linkis to version 1.3.2.

Deserialization of Untrusted Data in PublicCMS v.4.0.202302.e allows a remote attacker to execute arbitrary code via a crafted script to the writeReplace function.

The Java Remote Management Interface of all versions of SVI MS Management System was discovered to contain a vulnerability due to insecure deserialization of user-supplied content, which can allow attackers to execute arbitrary code via a crafted serialized Java object.

A JNDI rebind operation in the default ORB listener in Payara Server 4.1.2.191 (Enterprise), 5.20.0 and newer (Enterprise), and 5.2020.1 and newer (Community), when Java 1.8u181 and earlier is used, allows remote attackers to load malicious code on the server once a JNDI directory scan is performed.

Deserialization of Untrusted Data vulnerability in Liquid Web GiveWP allows Object Injection.This issue affects GiveWP: from n/a through 3.14.1.

A Java insecure deserialization vulnerability in Adobe LiveCycle ES4 version 11.0 and earlier allows unauthenticated remote attackers to gain operating system code execution by submitting specially crafted Java serialized objects to a specific URL. Adobe LiveCycle ES4 version 11.0.1 and later may be vulnerable if the application is installed with Java environment 7u21 and earlier. Exploitation of the vulnerability depends on two factors: insecure deserialization methods used in the Adobe LiveCycle application, and the use of Java environments 7u21 and earlier. The code execution is performed in the context of the account that is running the Adobe LiveCycle application. If the account is privileged, exploitation provides privileged access to the operating system. NOTE: This vulnerability only affects products that are no longer supported by the maintainer.

The Java Remote Management Interface of all versions of Orlansoft ERP was discovered to contain a vulnerability due to insecure deserialization of user-supplied content, which can allow attackers to execute arbitrary code via a crafted serialized Java object.

A accessmgrservlet classname deserialization of untrusted data remote code execution vulnerability was discovered in HPE Intelligent Management Center (iMC) version(s): Prior to iMC PLAT 7.3 (E0705P07).

The Lead Generated WordPress Plugin, version <= 1.23, was affected by an unauthenticated insecure deserialization issue. The tve_labels parameter of the tve_api_form_submit action is passed to the PHP unserialize() function without being sanitized or verified, and as a result could lead to PHP object injection, which when combined with certain class implementations / gadget chains could be leveraged to perform a variety of malicious actions granted a POP chain is also present.

There is a vulnerability caused by unsafe Java deserialization that allows for arbitrary command execution in a containerized environment within Airwave Glass before 1.3.3. Successful exploitation can lead to complete compromise of the underlying host operating system.

Deserialization of Untrusted Data vulnerability in Rocketgenius Inc. Gravity Forms.This issue affects Gravity Forms: from n/a through 2.7.3.

An insecure unserialize vulnerability was discovered in ThinkAdmin versions 4.x through 6.x in app/admin/controller/api/Update.php and app/wechat/controller/api/Push.php, which may lead to arbitrary remote code execution.

Remote Code Execution has been discovered in OpenText™ iManager 3.2.6.0200. The vulnerability can trigger command injection and insecure deserialization issues.

Snappy is a PHP library allowing thumbnail, snapshot or PDF generation from a url or a html page. Prior to version 1.4.2, Snappy is vulnerable to PHAR deserialization due to a lack of checking on the protocol before passing it into the `file_exists()` function. If an attacker can upload files of any type to the server he can pass in the phar:// protocol to unserialize the uploaded file and instantiate arbitrary PHP objects. This can lead to remote code execution especially when snappy is used with frameworks with documented POP chains like Laravel/Symfony vulnerable developer code. If a user can control the output file from the `generateFromHtml()` function, it will invoke deserialization. This vulnerability is capable of remote code execution if Snappy is used with frameworks or developer code with vulnerable POP chains. It has been fixed in version 1.4.2.

This vulnerability can affect all Dubbo users stay on version 2.7.6 or lower. An attacker can send RPC requests with unrecognized service name or method name along with some malicious parameter payloads. When the malicious parameter is deserialized, it will execute some malicious code. More details can be found below.

In Apache ShardingSphere(incubator) 4.0.0-RC3 and 4.0.0, the ShardingSphere's web console uses the SnakeYAML library for parsing YAML inputs to load datasource configuration. SnakeYAML allows to unmarshal data to a Java type By using the YAML tag. Unmarshalling untrusted data can lead to security flaws of RCE.

Pivotal Spring Framework through 5.3.16 suffers from a potential remote code execution (RCE) issue if used for Java deserialization of untrusted data. Depending on how the library is implemented within a product, this issue may or not occur, and authentication may be required. NOTE: the vendor's position is that untrusted data is not an intended use case. The product's behavior will not be changed because some users rely on deserialization of trusted data.

It was noticed that Apache Heron 0.20.2-incubating, Release 0.20.1-incubating, and Release v-0.20.0-incubating does not configure its YAML parser to prevent the instantiation of arbitrary types, resulting in a remote code execution vulnerabilities (CWE-502: Deserialization of Untrusted Data).

Deserialization of Untrusted Data vulnerability in 8theme XStore Core.This issue affects XStore Core: from n/a through 5.3.5.

QuantConnect Lean versions from 2.3.0.0 to 2.4.0.1 are affected by an insecure deserialization vulnerability due to insecure configuration of TypeNameHandling property in Json.NET library.

jsonpickle through 1.4.1 allows remote code execution during deserialization of a malicious payload through the decode() function. Note: It has been argued that this is expected and clearly documented behaviour. pickle is known to be capable of causing arbitrary code execution, and must not be used with un-trusted data

GPT Academic provides interactive interfaces for large language models. A vulnerability was found in gpt_academic versions 3.64 through 3.73. The server deserializes untrustworthy data from the client, which may risk remote code execution. Any device that exposes the GPT Academic service to the Internet is vulnerable. Version 3.74 contains a patch for the issue. There are no known workarounds aside from upgrading to a patched version.

Deserialization of Untrusted Data vulnerability in Repute Infosystems ARMember.This issue affects ARMember: from n/a through 4.0.26.

Retrieval-based-Voice-Conversion-WebUI is a voice changing framework based on VITS. Versions 2.2.231006 and prior are vulnerable to unsafe deserialization. The ckpt_dir variable takes user input (e.g. a path to a model) and passes it to the change_info function in export.py, which uses it to load the model on that path with torch.load, which can lead to unsafe deserialization and remote code execution. As of time of publication, no known patches exist.