LiteLLM is a proxy server (AI Gateway) to call LLM APIs in OpenAI (or native) format. Prior to 1.83.0, the /config/update endpoint does not enforce admin role authorization. A user who is already authenticated into the platform can then use this endpoint to modify proxy configuration and environment variables, register custom pass-through endpoint handlers pointing to attacker-controlled Python code, achieving remote code execution, read arbitrary server files by setting UI_LOGO_PATH and fetching via /get_image, and take over other privileged accounts by overwriting UI_USERNAME and UI_PASSWORD environment variables. Fixed in v1.83.0.
A flaw was found in libssh. On servers with GSSAPIKeyExchange enabled, the gssapi-keyex path does not verify whether the authenticated Kerberos principal is authorized for the requested local user, allowing authenticated clients to log in as arbitrary users.
An issue was discovered in OpenStack Keystone before 29.0.2. When combined with an application credential impersonation vulnerability, an attacker with the member role on a project can escalate to admin by chaining unrestricted application credentials with Keystone trusts. The impersonated token carries the victim's identity, which passes the trustor validation check. Keystone then validates the delegated roles against the victim's actual role assignments in the database, not the roles on the requesting token. This allows the attacker to create a trust delegating the victim's admin role to themselves. The trust persists independently, and additional trusts and application credentials can be created to maintain access. All actions are logged under the victim's identity.
LiteLLM prior to 1.83.10 allows a user to modify their own user_role via the /user/update endpoint. While the endpoint correctly restricts users to updating only their own account, it does not restrict which fields may be changed. A user who can reach this endpoint can set their role to proxy_admin, gaining full administrative access to LiteLLM including all users, teams, keys, models, and prompt history. Users with the org_admin role have legitimate access to this endpoint and can exploit this vulnerability without chaining any additional flaw.
LiteLLM prior to 1.83.14 allows an authenticated internal_user to create API keys with access to routes that their role does not permit. When generating a key, the allowed_routes field is stored without verifying that the specified routes fall within the user's own permissions. A key created with access to admin-only routes can then be used to reach those routes successfully, bypassing the role-based access controls that would otherwise block the request, enabling full privilege escalation from internal_user to proxy_admin.
An issue was discovered in OpenStack Keystone before 29.0.2. The Keystone RBAC policy enforcer in enforce_call unconditionally merges the raw JSON request body into the policy enforcement dictionary via policy_dict.update(json_input.copy()), overwriting trusted target data that was previously set from database lookups. Because flask.request.get_json is called with force=True, this works regardless of Content-Type or HTTP method. Any authenticated user can inject arbitrary policy target attributes (e.g., user_id, project_id) into the request body to bypass RBAC checks and perform unauthorized operations on resources belonging to other users or projects. This was introduced in commit 5ea59f52 (Rocky/14.0.0).
A flaw was found in Keycloak in versions from 12.0.0 and before 15.1.1 which allows an attacker with any existing user account to create new default user accounts via the administrative REST API even when new user registration is disabled.
A flaw was found in Wildfly Security Manager, running under JDK 11 or 8, that authorized requests for any requester. This flaw could be used by a malicious app deployed on the app server to access unauthorized information and possibly conduct further attacks. Versions shipped with Red Hat Jboss EAP 7 and Red Hat SSO 7 are vulnerable to this issue.
Spring Framework version 5.0.5 when used in combination with any versions of Spring Security contains an authorization bypass when using method security. An unauthorized malicious user can gain unauthorized access to methods that should be restricted.
Paramiko version 2.4.1, 2.3.2, 2.2.3, 2.1.5, 2.0.8, 1.18.5, 1.17.6 contains a Incorrect Access Control vulnerability in SSH server that can result in RCE. This attack appear to be exploitable via network connectivity.
JupyterLab is an extensible environment for interactive and reproducible computing, based on the Jupyter Notebook Architecture. From 4.0.0 to 4.5.6, the allow-list of extensions that can be installed from PyPI Extension Manager (allowed_extensions_uris) is not correctly enforced by JupyterLab. The PyPI Extension Manager was not contained to packages listed on the default PyPI index. This vulnerability is fixed in 4.5.7.
A flaw was found in the ptp4l program of the linuxptp package. A missing length check when forwarding a PTP message between ports allows a remote attacker to cause an information leak, crash, or potentially remote code execution. The highest threat from this vulnerability is to data confidentiality and integrity as well as system availability. This flaw affects linuxptp versions before 3.1.1, before 2.0.1, before 1.9.3, before 1.8.1, before 1.7.1, before 1.6.1 and before 1.5.1.
An incorrect access control flaw was found in the kiali-operator in versions before 1.33.0 and before 1.24.7. This flaw allows an attacker with a basic level of access to the cluster (to deploy a kiali operand) to use this vulnerability and deploy a given image to anywhere in the cluster, potentially gaining access to privileged service account tokens. The highest threat from this vulnerability is to data confidentiality and integrity as well as system availability.
A flaw was found in noobaa-operator in versions before 5.7.0, where internal RPC AuthTokens between the noobaa operator and the noobaa core are leaked into log files. An attacker with access to the log files could use this AuthToken to gain additional access into noobaa deployment and can read/modify system configuration.
The Gluster file system through versions 4.1.4 and 3.12 is vulnerable to a heap-based buffer overflow in the '__server_getspec' function via the 'gf_getspec_req' RPC message. A remote authenticated attacker could exploit this to cause a denial of service or other potential unspecified impact.
A flaw was found in postgresql in versions before 13.3, before 12.7, before 11.12, before 10.17 and before 9.6.22. While modifying certain SQL array values, missing bounds checks let authenticated database users write arbitrary bytes to a wide area of server memory. The highest threat from this vulnerability is to data confidentiality and integrity as well as system availability.
It was found that glusterfs server does not properly sanitize file paths in the "trusted.io-stats-dump" extended attribute which is used by the "debug/io-stats" translator. Attacker can use this flaw to create files and execute arbitrary code. To exploit this attacker would require sufficient access to modify the extended attributes of files on a gluster volume.
A flaw was found in RPC request using gfs2_create_req in glusterfs server. An authenticated attacker could use this flaw to create arbitrary files and execute arbitrary code on glusterfs server nodes.
A privilege escalation vulnerability exists in the oslo.privsep functionality of OpenStack git master 05194e7618 and prior. Overly permissive functionality within tools leveraging this library within a container can lead increased privileges.
In Mercurial before 4.1.3, "hg serve --stdio" allows remote authenticated users to launch the Python debugger, and consequently execute arbitrary code, by using --debugger as a repository name.
Math.js is an extensive math library for JavaScript and Node.js. From version 13.1.0 to before version 15.2.0, arbitrary JavaScript can be executed via the expression parser of mathjs. This issue has been patched in version 15.2.0.
Improper Input Validation, Improper Control of Generation of Code ('Code Injection') vulnerability in Apache ActiveMQ, Apache ActiveMQ Broker, Apache ActiveMQ All. An authenticated attacker can use the admin web console page to construct a malicious broker name that bypasses name validation to include an xbean binding that can be later used by a VM transport to load a remote Spring XML application. The attacker can then use the DestinationView mbean to send a message to trigger a VM transport creation that will reference this malicious broker name which can lead to loading the malicious Spring XML context file. Because Spring's ResourceXmlApplicationContext instantiates all singleton beans before the BrokerService validates the configuration, arbitrary code execution occurs on the broker's JVM through bean factory methods such as Runtime.exec(). This issue affects Apache ActiveMQ: before 5.19.6, from 6.0.0 before 6.2.5; Apache ActiveMQ Broker: before 5.19.6, from 6.0.0 before 6.2.5; Apache ActiveMQ All: before 5.19.6, from 6.0.0 before 6.2.5. Users are recommended to upgrade to version 6.2.5 or 5.19.6, which fixes the issue.
In ConsoleKit before 0.4.2, an intended security policy restriction bypass was found. This flaw allows an authenticated system user to escalate their privileges by initiating a remote VNC session.
A privilege escalation flaw was found in openshift4/ose-docker-builder. The build container runs with high privileges using a chrooted environment instead of runc. If an attacker can gain access to this build container, they can potentially utilize the raw devices of the underlying node, such as the network and storage devices, to at least escalate their privileges to that of the cluster admin. The highest threat from this vulnerability is to data confidentiality and integrity as well as system availability.
When an SSH server authentication callback returned PartialSuccessError with non-nil Permissions, those permissions were silently discarded, potentially dropping certificate restrictions such as force-command after a second factor succeeded. Returning non-nil Permissions with PartialSuccessError now results in a connection error.
In ovirt-engine 4.1, if a host was provisioned with cloud-init, the root password could be revealed through the REST interface.
Improper Input Validation, Improper Control of Generation of Code ('Code Injection') vulnerability in Apache ActiveMQ Broker, Apache ActiveMQ All, Apache ActiveMQ. An authenticated attacker may bypass the fix in CVE-2026-34197 by adding a connector using an HTTP Discovery transport via BrokerView.addNetworkConnector or BrokerView.addConnector through Jolokia if the activemq-http module is on the classpath. A malicious HTTP endpoint can return a VM transport through the HTTP URI which will bypass the validation added in CVE-2026-34197. The attacker can then use the VM transport's brokerConfig parameter to load a remote Spring XML application context using ResourceXmlApplicationContext. Because Spring's ResourceXmlApplicationContext instantiates all singleton beans before the BrokerService validates the configuration, arbitrary code execution occurs on the broker's JVM through bean factory methods such as Runtime.exec(). This issue affects Apache ActiveMQ Broker: before 5.19.6, from 6.0.0 before 6.2.5; Apache ActiveMQ All: before 5.19.6, from 6.0.0 before 6.2.5; Apache ActiveMQ: before 5.19.6, from 6.0.0 before 6.2.5. Users are recommended to upgrade to version 5.19.6 or 6.2.5, which fixes the issue.
A use-after-free vulnerability was found in drivers/nvme/target/tcp.c` in `nvmet_tcp_free_crypto` due to a logical bug in the NVMe/TCP subsystem in the Linux kernel. This issue may allow a malicious user to cause a use-after-free and double-free problem, which may permit remote code execution or lead to local privilege escalation.
Jupyter Server is the backend for Jupyter web applications. In versions 2.17.0 and earlier, a path traversal vulnerability in the REST API allows an authenticated user to escape the configured root_dir and access sibling directories whose names begin with the same prefix as the root_dir. For example, with a root_dir named "test", the API permits access to a sibling directory named "testtest" through a crafted request to the /api/contents endpoint using encoded path components. An attacker can read, write, and delete files in affected sibling directories. Multi-tenant deployments using predictable naming schemes are particularly at risk, as a user with a directory named "user1" could access directories for user10 through user19 and beyond. A user who can choose a single-character folder name could gain access to a significant number of sibling directories. Version 2.18.0 contains a fix. As a workaround, ensure folder names do not share a common prefix with any sibling directory.
A critical vulnerability exists in the NLTK downloader component of nltk/nltk, affecting all versions. The _unzip_iter function in nltk/downloader.py uses zipfile.extractall() without performing path validation or security checks. This allows attackers to craft malicious zip packages that, when downloaded and extracted by NLTK, can execute arbitrary code. The vulnerability arises because NLTK assumes all downloaded packages are trusted and extracts them without validation. If a malicious package contains Python files, such as __init__.py, these files are executed automatically upon import, leading to remote code execution. This issue can result in full system compromise, including file system access, network access, and potential persistence mechanisms.
Traefik is an HTTP reverse proxy and load balancer. Prior to versions 2.11.42, 3.6.11, and 3.7.0-ea.3, when `headerField` is configured with a non-canonical HTTP header name (e.g., `x-auth-user` instead of `X-Auth-User`), an authenticated attacker can inject their own canonical version of that header to impersonate any identity to the backend. The backend receives two header entries — the attacker-injected canonical one is read first, overriding Traefik's non-canonical write. Versions 2.11.42, 3.6.11, and 3.7.0-ea.3 patch the issue.
A flaw was found in Red Hat Quay's handling of resumable container image layer uploads. The upload process stores intermediate data in the database using a format that, if tampered with, could allow an attacker to execute arbitrary code on the Quay server.
Jenkins 2.554 and earlier, LTS 2.541.2 and earlier does not safely handle symbolic links during the extraction of .tar and .tar.gz archives, allowing crafted archives to write files to arbitrary locations on the filesystem, restricted only by file system access permissions of the user running Jenkins. This can be exploited to deploy malicious scripts or plugins on the controller by attackers with Item/Configure permission, or able to control agent processes.
The Adversarial Robustness Toolbox (ART) thru 1.20.1 contains a remote code execution vulnerability in its Kubeflow component. The robustness evaluation function for PyTorch models uses the unsafe eval() function to dynamically evaluate user-supplied strings for the LossFn and Optimizer parameters without any sanitization or security restrictions. An attacker can exploit this by providing a specially crafted string that contains arbitrary Python code, which will be executed when eval() is called, leading to complete compromise of the system running the ART evaluation.
Immutable.js provides many Persistent Immutable data structures. Prior to versions 3.8.3, 4.3.7, and 5.1.5, Prototype Pollution is possible in immutable via the mergeDeep(), mergeDeepWith(), merge(), Map.toJS(), and Map.toObject() APIs. This issue has been patched in versions 3.8.3, 4.3.7, and 5.1.5.
A flaw was found in org.keycloak.broker.saml. When a disabled Security Assertion Markup Language (SAML) client is configured as an Identity Provider (IdP)-initiated broker landing target, it can still complete the login process and establish a Single Sign-On (SSO) session. This allows a remote attacker to gain unauthorized access to other enabled clients without re-authentication, effectively bypassing security restrictions.
`simple-git`, an interface for running git commands in any node.js application, has an issue in versions 3.15.0 through 3.32.2 that allows an attacker to bypass two prior CVE fixes (CVE-2022-25860 and CVE-2022-25912) and achieve full remote code execution on the host machine. Version 3.23.0 contains an updated fix for the vulnerability.
The ConsulRegistry in the camel-consul component (class org.apache.camel.component.consul.ConsulRegistry and its inner ConsulRegistryUtils.deserialize method) read Java-serialized values from the Consul KV store and passed them to ObjectInputStream.readObject() without configuring an ObjectInputFilter. An attacker who can write to the Consul KV store backing a Camel ConsulRegistry instance could inject a malicious serialized Java object that is deserialized the next time Camel performs a lookup against that registry, leading to arbitrary code execution in the Camel process. The issue mirrors the class of vulnerability already addressed for other Camel components in CVE-2024-22369, CVE-2024-23114 and CVE-2026-25747, and was overlooked during the original remediation of those CVEs. This issue affects Apache Camel: from 3.0.0 before 4.14.6, from 4.15.0 before 4.18.1. Users are recommended to upgrade to version 4.19.0, which fixes the issue. If users are on the 4.14.x LTS releases stream, then they are suggested to upgrade to 4.14.6. If users are on the 4.18.x releases stream, then they are suggested to upgrade to 4.18.1.
A flaw was found in the integration of Active Directory and the System Security Services Daemon (SSSD) on Linux systems. In default configurations, the Kerberos local authentication plugin (sssd_krb5_localauth_plugin) is enabled, but a fallback to the an2ln plugin is possible. This fallback allows an attacker with permission to modify certain AD attributes (such as userPrincipalName or samAccountName) to impersonate privileged users, potentially resulting in unauthorized access or privilege escalation on domain-joined Linux hosts.
A flaw was found in Keycloak. An authenticated user with low privileges can exploit this vulnerability by sending an oversized subject_token JSON Web Token (JWT) to the TokenEndpoint. When the token exceeds a 4000-character limit, it is silently dropped, causing the system to fall back to client credentials. This allows the user to gain the permissions of the client's service account, leading to privilege escalation.
Redis is an in-memory data structure store. In versions of redis-server up to 8.6.3, the RESTORE command does not properly validate serialized values. An authenticated attacker with permission to execute RESTORE can supply a crafted serialized payload that triggers invalid memory access and may lead to remote code execution. A workaround is to restrict access to the RESTORE command with ACL rules. This is patched in version 8.6.3.
Impact: When using Socks5ProxyAgent, undici reuses a single connection pool across different origins without verifying that the pool's origin matches the requested origin. All requests are dispatched through the pool connected to the first origin, regardless of the intended destination. This causes cross-origin request routing: credentials and request data intended for origin B are sent to origin A, responses from the wrong origin are trusted, and HTTPS requests may be silently downgraded to HTTP. Impacted users are applications that use Socks5ProxyAgent (directly or via setGlobalDispatcher) and make requests to more than one origin. This was introduced in undici 7.23.0 via PR #4385 and affects all versions through 8.1.0. Patches: Upgrade to undici v7.26.0 or v8.2.0. Workarounds: Use a separate Socks5ProxyAgent instance per origin, or avoid using Socks5ProxyAgent with multiple origins.
Integer wraparound in multiple PostgreSQL server features allows an unprivileged database user to cause the server to undersize an allocation and write out-of-bounds. This may execute arbitrary code as the operating system user running the database. In applications that pass gigabyte-scale user inputs to the relevant database functions, the application input provider may achieve a segmentation fault. Versions before PostgreSQL 18.4, 17.10, 16.14, 15.18, and 14.23 are affected.
A vulnerability was found in OpenShift AI that allows for authentication bypass and privilege escalation across models within the same namespace. When deploying AI models, the UI provides the option to protect models with authentication. However, credentials from one model can be used to access other models and APIs within the same namespace. The exposed ServiceAccount tokens, visible in the UI, can be utilized with oc --token={token} to exploit the elevated view privileges associated with the ServiceAccount, leading to unauthorized access to additional resources.
In Jenkins 2.567 and earlier, LTS 2.555.2 and earlier, it is possible for attackers to have Jenkins deserialize arbitrary types defined in Jenkins core or plugins from an attacker-controlled `config.xml` submission in a way that allows them to handle HTTP requests afterwards. This can be used to impersonate any user and send HTTP requests on their behalf, up to and including use of the Script Console to run arbitrary code, or to read arbitrary files from the Jenkins controller.
A further incomplete fix for a previous advisory CVE-2026-44417 (Untrusted JMS configuration can lead to RCE) for Apache CXF has been identified, which can allow code execution capabilities, if untrusted users are allowed to configure JMS for Apache CXF. Users are recommended to upgrade to versions 4.2.2 or 4.1.7, which fixes this issue.
LiteLLM is a proxy server (AI Gateway) to call LLM APIs in OpenAI (or native) format. From version 1.80.5 to before version 1.83.7, the POST /prompts/test endpoint accepted user-supplied prompt templates and rendered them without sandboxing. A crafted template could run arbitrary code inside the LiteLLM Proxy process. The endpoint only checks that the caller presents a valid proxy API key, so any authenticated user could reach it. Depending on how the proxy is deployed, this could expose secrets in the process environment (such as provider API keys or database credentials) and allow commands to be run on the host. This issue has been patched in version 1.83.7.
LiteLLM is a proxy server (AI Gateway) to call LLM APIs in OpenAI (or native) format. From version 1.74.2 to before version 1.83.7, two endpoints used to preview an MCP server before saving it — POST /mcp-rest/test/connection and POST /mcp-rest/test/tools/list — accepted a full server configuration in the request body, including the command, args, and env fields used by the stdio transport. When called with a stdio configuration, the endpoints attempted to connect, which spawned the supplied command as a subprocess on the proxy host with the privileges of the proxy process. The endpoints were gated only by a valid proxy API key, with no role check. Any authenticated user — including holders of low-privilege internal-user keys — could therefore run arbitrary commands on the host. This issue has been patched in version 1.83.7.
The camel-mina component's MinaConverter.toObjectInput(IoBuffer) type converter wraps an IoBuffer in a java.io.ObjectInputStream without applying any ObjectInputFilter or class-loading restrictions. When a Camel route uses camel-mina as a TCP or UDP consumer and requests conversion to ObjectInput (for example via getBody(ObjectInput.class) or @Body ObjectInput), an attacker sending a crafted serialized Java object over the network to the MINA consumer port can trigger arbitrary code execution in the context of the application during readObject(). This issue affects Apache Camel: from 3.0.0 before 4.14.6, from 4.15.0 before 4.18.2, from 4.19.0 before 4.20.0. Users are recommended to upgrade to version 4.20.0, which fixes the issue. If users are on the 4.14.x LTS releases stream, then they are suggested to upgrade to 4.14.6. If users are on the 4.18.x releases stream, then they are suggested to upgrade to 4.18.2.
The Adversarial Robustness Toolbox (ART) thru 1.20.1 contains a command-line argument injection vulnerability in its Kubeflow component (robustness_evaluation_fgsm_pytorch.py). The script uses the unsafe eval() function to parse string values provided via the --clip_values and --input_shape command-line arguments. This allows an attacker to inject arbitrary Python code into these arguments, which will be executed when eval() is called. The vulnerability can be exploited remotely if an attacker can control these arguments (e.g., through pipeline configuration or automated scripts), leading to arbitrary code execution on the system running the ART evaluation.