CVE Database

Kyverno is a policy engine designed for cloud native platform engineering teams. Prior to versions 1.17.2 and 1.16.4, an unchecked type assertion in the `forEach` mutation handler allows any user with permission to create a `Policy` or `ClusterPolicy` to crash the cluster-wide background controller into a persistent CrashLoopBackOff. The same bug also causes the admission controller to drop connections and block all matching resource operations. The crash loop persists until the policy is deleted. The vulnerability is confined to the legacy engine, and CEL-based policies are unaffected. Versions 1.17.2 and 1.16.4 fix the issue.

Due to an uncontrolled resource consumption (Denial of Service) vulnerability, an authenticated attacker with regular user privileges and network access can repeatedly invoke a remote-enabled function module with an excessively large loop-control parameter. This triggers prolonged loop execution that consumes excessive system resources, potentially rendering the system unavailable. Successful exploitation results in a denial-of-service condition that impacts availability, while confidentiality and integrity remain unaffected.

OpenObserve is a cloud-native observability platform. In 0.70.3 and earlier, the validate_enrichment_url function in src/handler/http/request/enrichment_table/mod.rs fails to block IPv6 addresses because Rust's url crate returns them with surrounding brackets (e.g. "[::1]" not "::1"). An authenticated attacker can reach internal services blocked from external access. On cloud deployments this enables retrieval of IAM credentials via AWS IMDSv1 (169.254.169.254), GCP metadata, or Azure IMDS. On self-hosted deployments it allows probing internal network services.

barebox prior to version 2026.04.0 contains out-of-bounds read vulnerabilities in ext4 extent parsing due to missing validation of the eh_entries field against buffer capacity in fs/ext4/ext4_common.c. Attackers can supply a malicious ext4 filesystem image via USB, SD card, or network boot to trigger heap out-of-bounds reads during boot-time filesystem parsing, potentially redirecting reads to arbitrary disk offsets.

AutoGPT is a workflow automation platform for creating, deploying, and managing continuous artificial intelligence agents. In versions prior to 0.6.52, the Fill Text Template block is vulnerable to a Denial of Service (DoS) attack. While the backend implements a SandboxedEnvironment to prevent unauthorized attribute access (e.g., blocking __class__), it fails to limit the computational complexity or execution time of the expressions. An attacker can input computationally expensive Python/Jinja2 expressions that consume the server's CPU and memory, leading to a complete system hang or crash. In multi-tenant or self-hosted environments, this results in a complete service outage and "noisy neighbor" effects that require manual administrative intervention to recover. This issue has been fixed in version 0.6.52.

OpenClaw before 2026.4.14 contains an improper access control vulnerability in browser snapshot, screenshot, and tab routes that fail to consistently validate the final browser target after navigation. Authenticated callers can bypass SSRF restrictions to expose internal or disallowed page content by exploiting route-driven navigation without proper policy re-validation.

OpenClaw before 2026.3.24 contains a path traversal vulnerability in sandbox enforcement allowing sandboxed agents to read arbitrary files from other agents' workspaces via unnormalized mediaUrl or fileUrl parameter keys. Attackers can exploit incomplete parameter validation in normalizeSandboxMediaParams and missing mediaLocalRoots context to access sensitive files including API keys and configuration data outside designated sandbox roots.

PraisonAIAgents is a multi-agent teams system. Prior to 1.5.128, the web_crawl() function in praisonaiagents/tools/web_crawl_tools.py accepts arbitrary URLs from AI agents with zero validation. No scheme allowlisting, hostname/IP blocklisting, or private network checks are applied before fetching. This allows an attacker (or prompt injection in crawled content) to force the agent to fetch cloud metadata endpoints, internal services, or local files via file:// URLs. This vulnerability is fixed in 1.5.128.

PraisonAI is a multi-agent teams system. Prior to version 4.5.90, passthrough() and apassthrough() in praisonai accept a caller-controlled api_base parameter that is concatenated with endpoint and passed directly to httpx.Client.request() when the litellm primary path raises AttributeError. No URL scheme validation, private IP filtering, or domain allowlist is applied, allowing requests to any host reachable from the server. This issue has been patched in version 4.5.90.

arduino-TuyaOpen before version 1.2.1 contains an out-of-bounds memory read vulnerability in the TuyaIoT component. An attacker who hijacks or controls the Tuya cloud service can issue malicious DP event data to victim devices, causing out-of-bounds memory access that may result in information disclosure or a denial-of-service condition.

Host and event action script input is validated with a regex (set by the administrator), but the validation runs in multiline mode. If ^ and $ anchors are used in user input validation, an injected newline lets authenticated users bypass the check and inject shell commands.

Outline is a service that allows for collaborative documentation. The `shares.create` API endpoint starting in version 0.86.0 and prior to version 1.7.0 has an insecure direct object reference.. When both `collectionId` and `documentId` are provided in the request, the authorization logic only checks access to the collection, completely ignoring the document. This allows an authenticated attacker to generate a valid public share link for any document on the platform, including documents belonging to other workspaces. The full document contents can then be retrieved via the `documents.info` endpoint. Version 1.7.0 contains a patch.

Kyverno is a policy engine designed for cloud native platform engineering teams. The patch for CVE-2026-22039 fixed cross-namespace privilege escalation in Kyverno's `apiCall` context by validating the `URLPath` field. However, the ConfigMap context loader has the identical vulnerability — the `configMap.namespace` field accepts any namespace with zero validation, allowing a namespace admin to read ConfigMaps from any namespace using Kyverno's privileged service account. This is a complete RBAC bypass in multi-tenant Kubernetes clusters. An updated fix is available in version 1.17.2.

Nexxt Solutions Nebula 300+ firmware through version 12.01.01.37 does not implement rate limiting or account lockout mechanisms on authentication interfaces. An attacker can perform unlimited authentication attempts against endpoints that rely on credential validation, enabling brute-force attacks to guess administrative credentials without restriction.

OpenReplay is a self-hosted session replay suite. Prior to 1.26.0, OpenReplay's Python API exposes several app_apikey routes that trust a caller-provided projectKey after validating only that the API key itself is valid and that the target projectKey exists. The authorization flow does not verify that the authenticated API key and the requested project belong to the same tenant. Because the public tracker design exposes projectKey to browser-side code, an attacker who owns any valid API key for their own tenant can target another tenant's project by reusing that public projectKey. The vulnerable routes allow the attacker to enumerate victim user sessions and then retrieve sensitive session event data across the tenant boundary. This vulnerability is fixed in 1.26.0.

In multiple functions of TaskFragmentOrganizerController.java, there is a possible activity token leak due to a logic error in the code. This could lead to local escalation of privilege with no additional execution privileges needed. User interaction is not needed for exploitation.

OpenClaw before 2026.4.10 contains an incomplete navigation guard vulnerability that allows attackers to trigger navigation without complete SSRF policy enforcement. Browser press/type style interactions, including pressKey and type submit flows, can bypass post-action security checks to execute unauthorized navigation.

A flaw was found in Keycloak. Keycloak's Security Assertion Markup Language (SAML) broker endpoint does not properly validate encrypted assertions when the overall SAML response is not signed. An attacker with a valid signed SAML assertion can exploit this by crafting a malicious SAML response. This allows the attacker to inject an encrypted assertion for an arbitrary principal, leading to unauthorized access and potential information disclosure.

WWBN AVideo is an open source video platform. In versions up to and including 29.0, an unauthenticated user can read APISecret from objects/plugins.json.php and use it to call protected API endpoints (e.g. users_list) without logging in. Commit 1c36f229d0a103528fb9f64d0a1cc0e1e8f5999b contains an updated fix.

FastGPT is a knowledge-based AI application platform. Prior to 4.15.0-beta4, the HTTP-tool OpenAPI schema importer validates only the top-level URL before passing it to SwaggerParser.bundle, whose remote reference resolver fetches $ref URLs without FastGPT's internal-address guard and returns fetched content inline, allowing an authenticated team member to read internal services or cloud metadata. This issue is fixed in version 4.15.0-beta4.

Affected devices do not properly restrict access to the web browser via the Control Panel when no corresponding security mechanisms are in place. This could allow an unauthenticated attacker to gain unauthorized access to the web browser, potentially enabling the discovery of backdoors, performing unauthorized actions, or exploiting misconfigurations that may lead to further system compromise.

LibreChat is a ChatGPT clone with additional features. Versions 0.8.2-rc2 through 0.8.2 are vulnerable to a server-side request forgery (SSRF) attack when using agent actions or MCP. Although a previous SSRF vulnerability (https://github.com/danny-avila/LibreChat/security/advisories/GHSA-rgjq-4q58-m3q8) was reported and patched, the fix only introduced hostname validation. It does not verify whether DNS resolution results in a private IP address. As a result, an attacker can still bypass the protection and gain access to internal resources, such as an internal RAG API or cloud instance metadata endpoints. Version 0.8.3-rc1 contains a patch.

In OpenStack Keystone before 28.0.1, the LDAP identity backend does not convert the user enabled attribute to a boolean when the user_enabled_invert configuration option is False (the default). The _ldap_res_to_model method in the UserApi class only performed string-to-boolean conversion when user_enabled_invert was True. When False, the raw string value from LDAP (e.g., "FALSE") was used directly. Since non-empty strings are truthy in Python, users marked as disabled in LDAP were treated as enabled by Keystone, allowing them to authenticate and perform actions. All deployments using the LDAP identity backend without user_enabled_invert=True or user_enabled_emulation are affected.

Daytona is a secure and elastic infrastructure runtime for AI-generated code execution and agent workflows. Prior to 0.185.0, Daytona's organization role update and delete endpoints authorized the caller as an owner of the organization named in the request path, but resolved and mutated the target role by its identifier alone, without verifying the role belonged to that organization. An authenticated user who owns any organization (organizations are self-service) could therefore modify the permissions of, or delete, a role belonging to a different organization, given that role's identifier. This vulnerability is fixed in 0.185.0.

An issue in Eufy Homebase 2 version 3.3.4.1h allows a local attacker to obtain sensitive information via the cryptographic scheme.

The Loki datasource plugin's callResource handler contains a path traversal vulnerability. An authenticated Viewer-role user can escape the plugin's resource sandbox and access administrative Loki endpoints (e.g. /config, /services, /ready) to extract sensitive backend configuration and internal service information.

OpenClaw before 2026.5.18 contains a server-side request forgery vulnerability in browser control that allows authenticated users to bypass private-network navigation checks through Playwright act interactions. Attackers can trigger navigation to private-network targets via action-triggered redirects and subsequently read restricted page content using browser evaluation capabilities.

Nhost is an open source Firebase alternative with GraphQL. Prior to version 1.41.0, The Nhost CLI MCP server, when explicitly configured to listen on a network port, applies no inbound authentication and does not enforce strict CORS. This allows a malicious website visited on the same machine to issue cross-origin requests to the MCP server and invoke privileged tools using the developer's locally configured credentials. This vulnerability requires two explicit, non-default configuration steps to be exploitable. The default nhost mcp start configuration is not affected. This issue has been patched in version 1.41.0.

Untrusted Search Path vulnerability in B&R Industrial Automation GmbH APROL. This issue affects APROL: before R 4.4-01P5.

Open WebUI is a self-hosted artificial intelligence platform designed to operate entirely offline. Prior to 0.6.5, through the HTML rendering view, scripts can be injected and executed. The frontend provides a function to visualize the HTML content of a current chat. The content is embedded in an iFrame with the allow-scripts allow-forms allow-same-origin sandbox directive. This means that the content is placed in a sandbox but with permission to execute scripts and access the parent’s data (e.g., local storage). As a result, only a few functions are restricted (e.g., displaying an alert box), but in effect, the sandbox attribute is largely nullified. This vulnerability is fixed in 0.6.5.

A web page that contains unusual GPU shader code is loaded into the GPU compiler process and can trigger a write out-of-bounds write crash in the GPU shader compiler library. On certain platforms, when the compiler process has system privileges this could enable further exploits on the device. An edge case using a very small value in GPU shader code can cause a segmentation fault in the GPU shader compiler due to am out-of-bounds write.

Plane is an an open-source project management tool. Prior to version 1.2.2, a Full Read Server-Side Request Forgery (SSRF) vulnerability has been identified in the "Add Link" feature. This flaw allows an authenticated attacker with general user privileges to send arbitrary GET requests to the internal network and exfiltrate the full response body. By exploiting this vulnerability, an attacker can steal sensitive data from internal services and cloud metadata endpoints. Version 1.2.2 fixes the issue.

Server-Side Request Forgery (CWE-918) in Kibana allows authenticated users with connector management privileges to bypass the operator-configured connection allowlist. By configuring a Webhook connector with a crafted target, an attacker can cause Kibana to issue outbound requests to destinations that the egress restriction controls were intended to block.

Warp is an agentic development environment. From 0.2024.03.12.08.02.stable_01 until 0.2026.05.06.15.42.stable_01, Warp contains an OS command injection vulnerability in the WSL URL-opening fallback. When Warp is running under WSL and cannot open a URL through wslview, it falls back to a Windows command processor path. A URL controlled through terminal output can reach that fallback when the user opens the link. This vulnerability is fixed in 0.2026.05.06.15.42.stable_01.

Budibase is an open-source low-code platform. Prior to 3.39.0, the OAuth2 token fetch function in packages/server/src/sdk/workspace/oauth2/utils.ts uses raw fetch(config.url) with no SSRF protection. The safe wrapper fetchWithBlacklist() exists in the same codebase and is used in every other outbound HTTP call (automation steps, plugin downloads, object store), but was not applied to the OAuth2 token endpoint. A user with BUILDER role can point the OAuth2 token URL to internal services (CouchDB, cloud metadata) to exfiltrate sensitive data. This vulnerability is fixed in 3.39.0.

Sales Representative Arbitrary File Deletion in Groundhogg <= 4.4 versions.

Wallos is an open-source, self-hostable personal subscription tracker. In versions 4.8.4 and prior, the incomplete SSRF fix in Wallos validates webhook URLs via gethostbyname() but passes the original hostname to cURL without CURLOPT_RESOLVE pinning on 10 of 11 outbound HTTP endpoints, leaving a DNS rebinding TOCTOU window. At time of publication, there are no publicly available patches.

Budibase is an open-source low-code platform. Prior to 3.38.3, removeSecrets at packages/server/src/sdk/workspace/datasources/datasources.ts masks only datasource config fields whose schema type is DatasourceFieldType.PASSWORD. The Snowflake integration types its privateKey field as SENSITIVE_LONGFORM, which the filter skips. GET /api/datasources/:datasourceId lives on authorizedRoutes guarded by PermissionType.TABLE + PermissionLevel.READ. An authenticated BASIC user with any app role and call the endpoint and receive the full Snowflake PEM in plaintext. This vulnerability is fixed in 3.38.3.

Hermes WebUI before version 0.51.296 contains a workspace boundary bypass vulnerability that allows authenticated attackers to circumvent blocked-root path checks by exploiting an early return in the SSH/remote terminal profile workspace resolution logic within _remote_terminal_workspace_candidate(). Attackers can configure a remote terminal working directory to a system directory such as /etc, causing the workspace resolution path to accept it as a trusted local workspace root before the _is_blocked_workspace_path() guard executes, enabling read access to local system files through workspace file-read helpers.

Argo Workflows is an open source container-native workflow engine for orchestrating parallel jobs on Kubernetes. From 3.6.5 to 4.0.4, an unchecked array index in the pod informer's podGCFromPod() function causes a controller-wide panic when a workflow pod carries a malformed workflows.argoproj.io/pod-gc-strategy annotation. Because the panic occurs inside an informer goroutine (outside the controller's recover() scope), it crashes the entire controller process. The poisoned pod persists across restarts, causing a crash loop that halts all workflow processing until the pod is manually deleted. This vulnerability is fixed in 4.0.5 and 3.7.14.

Incorrect Authorization (CWE-863) in Kibana can lead to information disclosure via Privilege Abuse (CAPEC-122). A user with limited Fleet privileges can exploit an internal API endpoint to retrieve sensitive configuration data, including private keys and authentication tokens, that should only be accessible to users with higher-level settings privileges. The endpoint composes its response by fetching full configuration objects and returning them directly, bypassing the authorization checks enforced by the dedicated settings APIs.

An Authenticated NoSQL Injection vulnerability found in UniFi Network Application could allow a malicious actor with authenticated access to the network to escalate privileges.

Execution with Unnecessary Privileges (CWE-250) in Kibana’s Fleet plugin debug route handlers can lead reading index data beyond their direct Elasticsearch RBAC scope via Privilege Abuse (CAPEC-122). This requires an authenticated Kibana user with Fleet sub-feature privileges (such as agents, agent policies, and settings management).

Grav is a file-based Web platform. Prior to 2.0.0-rc.2, the Twig sandbox allow-list permits any user with the admin.pages role to call config.toArray() from within a page body, dumping the entire merged site configuration — including all plugin secrets (SMTP passwords, AWS keys, OAuth client secrets, API tokens) — into the rendered HTML. No administrator privileges are required. This vulnerability is fixed in 2.0.0-rc.2.

Predictable bucket naming in Vertex AI Experiments in Google Cloud Vertex AI from version 1.21.0 up to (but not including) 1.133.0 on Google Cloud Platform allows an unauthenticated remote attacker to achieve cross-tenant remote code execution, model theft, and poisoning via pre-creating predictably named Cloud Storage buckets (Bucket Squatting). This vulnerability was patched and no customer action is needed.

A vulnerability in the rm utility of uutils coreutils allows a bypass of the --preserve-root protection. The implementation uses a path-string check rather than comparing device and inode numbers to identify the root directory. An attacker or accidental user can bypass this safeguard by using a symbolic link that resolves to the root directory (e.g., /tmp/rootlink -> /), potentially leading to the unintended recursive deletion of the entire root filesystem.

NVIDIA KAI Scheduler contains a vulnerability where an attacker could access API endpoints without authorization. A successful exploit of this vulnerability might lead to information disclosure.

SEPPmail Secure Email Gateway before version 15.0.3 allows an attacker to bypass subject sanitization and forge tags such as [signed OK].

The extension fails to properly reset the generated MFA code after successful authentication. This leads to a possible MFA bypass for future login attempts by providing an empty string as MFA code to the extensions MFA provider.

Kestra is an open-source, event-driven orchestration platform. Prior to 1.0.43 and 1.3.19, several Kestra API endpoints accept a kestra:// URI from the client and pass it through StorageInterface.parentTraversalGuard before reading the underlying file from the local storage backend. The guard only inspects the literal URI.toString(), so a URL-encoded .. written as %2E%2E slips through. The downstream code then calls URI.getPath(), which decodes %2E%2E back to .., and the resulting path is handed to Paths.get(...) without normalization. The OS resolves the .. segments at open(2) time, so an authenticated user with a single execution can read any file the Kestra process has access to on the host filesystem (/etc/passwd, mounted secrets, other tenants' execution outputs, etc.). This vulnerability is fixed in 1.0.43 and 1.3.19.