| CVE |
Vendors |
Products |
Updated |
CVSS v3.1 |
| A critical remote code execution vulnerability exists in all versions of the HuggingFace transformers library prior to version 5.3.0. The vulnerability allows an attacker to craft a malicious `config.json` file containing the `_attn_implementation_internal` field set to an attacker-controlled HuggingFace Hub repository ID. When a victim loads this model using the standard `AutoModelForCausalLM.from_pretrained()` API, the library downloads and executes arbitrary Python code from the attacker's repository with the victim's full OS privileges. This issue arises due to unfiltered deserialization of configuration attributes, insufficient sanitization of internal fields, and unsandboxed execution of downloaded kernels. The vulnerability bypasses the `trust_remote_code` security mechanism, is invisible to the victim, and exploits the standard documented usage pattern, making it particularly severe. Users are advised to upgrade to version 5.3.0 or later to mitigate this issue. |
| NVIDIA NVTabular contains a vulnerability where an attacker could cause improper deserialization of untrusted data. A successful exploit of this vulnerability might lead to code execution, data tampering and information disclosure. |
| NVIDIA NVTabular contains a vulnerability where an attacker could cause improper deserialization of untrusted data. A successful exploit of this vulnerability might lead to code execution, data tampering, and information disclosure. |
| IBM WebSphere Application Server 9.0, and 8.5 is vulnerable to potential remote code execution due to deserialization of untrusted data via JAX-WS endpoints with WS-Security. |
| IBM WebSphere Application Server 9.0, and 8.5 is affected by an improper validation of user-supplied data during deserialization using the SAML Web Single Sign-On component. This could result in remote code execution via a crafted HTTP request when combined with a suitable gadget chain. |
| NVIDIA Transformers4Rec for Linux contains a vulnerability where an attacker could cause improper deserialization of untrusted data. A successful exploit of this vulnerability might lead to code execution, data tampering, and information disclosure. |
| Concrete CMS below 9.5.2 is vulnerable to PHP Object Injection via unserialize() calls in the Workflow, Form block, and File/Set components that lack the allowed_classes restriction. An unauthenticated attacker may trigger arbitrary PHP object instantiation if a malicious serialized payload has been placed in the database. Thanks XananasX7 and Sanjorn Keeratirungsan (dizconnect) for both independently reporting. The Concrete CMS security team gave this vulnerability a CVSS v.4.0 score of 8.4 with vector CVSS:4.0/AV:L/AC:L/AT:N/PR:H/UI:N/VC:H/VI:H/VA:H/SC:N/SI:N/SA:N. |
| Mirasvit Full Page Cache Warmer for Magento 2 before version 1.11.12 contains a PHP object injection vulnerability that allows unauthenticated attackers to achieve remote code execution by supplying a crafted serialized PHP object in the CacheWarmer cookie. Attackers can exploit the unrestricted call to PHP's native unserialize() function combined with gadget chains available in Magento and its dependencies to execute arbitrary code on the server. |
| Jenkins LDAP Plugin 807.v7d7de30930cf and earlier deserializes data from LDAP referrals without validation. |
| dd-trace-java is a Datadog APM client for Java. In versions of dd-trace-java 0.40.0 through prior to 1.60.2, the RMI instrumentation registered a custom endpoint that deserialized incoming data without applying serialization filters. On JDK version 16 and earlier, an attacker with network access to a JMX or RMI port on an instrumented JVM could exploit this to potentially achieve remote code execution. All three of the following conditions must be true to exploit this vulnerability: First, dd-trace-java is attached as a Java agent (`-javaagent`) on Java 16 or earlier. Second, a JMX/RMI port has been explicitly configured via `-Dcom.sun.management.jmxremote.port` and is network-reachable, Third, a gadget-chain-compatible library is present on the classpath. For JDK >= 17, no action is required, but upgrading is strongly encouraged. For JDK >= 8u121 < JDK 17, upgrade to dd-trace-java version 1.60.3 or later. For JDK < 8u121 and earlier where serialization filters are not available, apply the workaround. The workaround is to set the following environment variable to disable the RMI integration: `DD_INTEGRATION_RMI_ENABLED=false`. |
| ZDRES-232: resolveProxyClass Not Overridden - acceptMatchers Filter Bypass via java.lang.reflect.Proxy
Assessment: Fully addressed.
When the serialised stream contains a TC_PROXYCLASSDESC (the marker
for a java.lang.reflect.Proxy ), JDK’s ObjectInputStream.readProxyDesc()
is
dispatched. JDK then calls the default
ObjectInputStream.resolveProxyClass(interfaces) implementation, which
performs Class.forName(intf, false, latestUserDefinedLoader()) for EACH
interface name and constructs the proxy class — bypassing the accepted
classes list .
ZDRES-233: Class.forName(name, initialize=true, classLoader) in
readClassDescriptor Triggers Static Initialiser of Allow-Listed Classes
Assessment: Fully addressed.
For ANY class on the allow-list, deserialising a stream that names it triggers the class’s
(static initialiser) BEFORE any instance is constructed. This means an
attacker who supplies a class name on the allow-list (e.g., the
developer wrote accept(“com.myapp.*") , attacker supplies
com.myapp.SomeClass ) causes <clinit> of SomeClass — and many
real-world classes have side-effecting static initialisers
Both issues have been fixed. |
| A bug in Apache Airflow's XCom PATCH endpoint `PATCH /api/v2/xcomEntries/{key}` allowed an authenticated UI/API user with XCom write permission on a Dag to set XCom entries under reserved key names (e.g. `return_value`) that the matching POST endpoint already validated against `FORBIDDEN_XCOM_KEYS`. The endpoint also accepted serialized payload shapes the triggerer's deserializer treats as code; combined, this allowed RCE on the triggerer when the affected task next deferred. Affects deployments where untrusted users have XCom write permission on Dags that defer to the triggerer. This is a fix-bypass of CVE-2026-33858: PR #64148 added the `FORBIDDEN_XCOM_KEYS` validator only on the POST/set path; the PATCH path was not covered. Users who already upgraded for CVE-2026-33858 should additionally upgrade to `apache-airflow` 3.2.2 or later to cover the PATCH-path bypass. |
| Apache Airflow's scheduler-side deadline-reference decoder (`SerializedCustomReference.deserialize_reference`) imported and dispatched arbitrary class paths drawn from DAG-author-controlled serialized state without an allowlist or plugin-registry gate. A DAG author whose code reaches the scheduler — the default on single-host deployments where the DAG bundle is importable from the scheduler process — could embed a custom `DeadlineReference` whose serialized form named an attacker-controlled module path, causing the scheduler to `import_string(...)` and instantiate that class with a live SQLAlchemy session attached. Affects deployments where DAG-author code is less trusted than the scheduler process. Users are advised to upgrade to `apache-airflow` 3.2.2 or later. |
| An issue in ESA AnomalyMatch before 1.3.1 allow attackers to execute arbitrary code via crafted model checkpoint files. The affected components load model files from session directories using torch.load() with unrestricted deserialization. |
| A Deserialization of Untrusted Data vulnerability affecting Teamwork Cloud from No Magic Release 2022x through No Magic Release 2026x and Magic Collaboration Studio from CATIA Magic Release 2022x through CATIA Magic Release 2026x could lead to an unauthenticated remote code execution. |
| Deserialization of Untrusted Data vulnerability in Elated-Themes Aperitif allows Object Injection.
This issue affects Aperitif: from n/a through 1.6. |
| Deserialization of Untrusted Data vulnerability in Elated-Themes Töbel allows Object Injection.
This issue affects Töbel: from n/a through 1.8.1. |
| A flaw was found in camel-infinispan. This vulnerability involves unsafe deserialization in the ProtoStream remote aggregation repository. A remote attacker with low privileges could exploit this by sending specially crafted data, leading to arbitrary code execution. This allows the attacker to gain full control over the affected system, impacting its confidentiality, integrity, and availability. |
| Deserialization of untrusted data vulnerability in QOS.CH Sarl logback logback-core (HardenedObjectInputStream (logback-core) modules) allows Object Injection, albeit heavily restricted.
More precisely, an attacker able to influence serialized data sent to
SimpleSocketServer or SimpleSSLSocketServer can instantiate Proxy objects.
Although deserialization is heavily restricted by HardenedObjectInputStream and no
practical way to achieve remote code execution or significant privilege
escalation has been identified, this issue constitutes a bypass of the
intended security restrictions.
This issue affects logback: through 1.5.33 inclusive. |
| LangSmith Client SDKs provide SDK's for interacting with the LangSmith platform. Prior to LangSmith SDK Python 0.8.0 and JS/TS 0.6.0, the LangSmith SDK's prompt pull methods (pull_prompt / pull_prompt_commit in Python, pullPrompt / pullPromptCommit in JS/TS) fetch and deserialize prompt manifests from the LangSmith Hub. These manifests may contain serialized LangChain objects and model configuration that affect runtime behavior. When pulling a public prompt by owner/name identifier, the manifest content is controlled by an external party, but prior versions of the SDK did not distinguish this from pulling a prompt within the caller's own organization. This vulnerability is fixed in LangSmith SDK Python 0.8.0 and JS/TS 0.6.0. |
