NocoBase versions up to 2.1.20 contain a server-side request forgery (SSRF) vulnerability that allows authenticated administrators to make arbitrary outbound HTTP requests. This can be exploited via workflow request nodes, custom request action buttons, or the AI plugin. The vulnerability enables attackers to access internal network resources, including loopback addresses, private IP ranges, and cloud instance metadata endpoints, potentially leading to internal network reconnaissance and retrieval of IAM role credentials. Version 2.1.18 introduced a warning when the SERVER_REQUEST_WHITELIST is not configured, but no fix is explicitly stated.
CVE-2026-59100 is a broken object level authorization vulnerability in LobeChat through version 2.2.9. It allows authenticated attackers to access and modify chat-group agent data of other users by supplying arbitrary group identifiers. Attackers can perform operations such as getGroupAgents, updateAgentInGroup, and removeAgentsFromGroup without proper user-scoped checks, enabling unauthorized reading, modification, and removal of agents in other users' chat groups.
LobeChat versions up to 2.2.9 have an authorization bypass vulnerability in the semantic search feature. Authenticated attackers can exploit missing user-identifier checks to access other users' data, including text content, file names, and metadata. This occurs through manipulation of victim file or knowledge-base identifiers in the chunk retrieval and chat knowledge-base paths.
CVE-2026-59096 is a high-severity vulnerability in dapr versions 1.17.0 and 1.18.0 involving an origin validation error in Dapr Sentry's OIDC discovery endpoint. The endpoint derives issuer and jwks_uri values from the request Host header, trusting an attacker-controlled X-Forwarded-Host header without validation when no allowed-hosts list is configured (the default). This allows a remote unauthenticated attacker to poison the discovery document, causing relying parties to fetch JSON Web Key Sets (JWKS) from an attacker-controlled server and accept attacker-signed JWTs. Exploitation requires the OIDC server to be enabled without a configured jwt-issuer or oidc-allowed-hosts.
LobeChat versions before 2.2.10-canary.18 contain a server-side request forgery (SSRF) vulnerability. Authenticated attackers can exploit this flaw by supplying crafted URLs to the skill import and topic cover update endpoints, which do not use the project's SSRF-safe fetch wrapper. This allows attackers to make internal HTTP requests to arbitrary URLs, including sensitive internal services such as cloud instance metadata endpoints, potentially disclosing internal service responses and cloud credentials.
Weaviate versions prior to 1.38.0 contain a privilege escalation vulnerability due to improper verification of permissions during RBAC role assignments. Specifically, users with limited assign_and_revoke_users or assign_and_revoke_groups permissions can assign themselves or others high-privilege roles, including the built-in admin role, without verifying that they hold the permissions contained in those roles. This flaw allows unauthorized escalation to full administrative control of the database.
LobeChat versions before 2.2.10-canary.15 contain a regular expression denial of service (ReDoS) vulnerability. Authenticated attackers can supply a malicious GitHub repository URL path with catastrophic-backtracking regex patterns during skill import. This causes synchronous execution of a complex regex in the findSkillMd function, blocking the Node.js event loop and denying service to all users for tens of seconds per request. The issue is fixed in version 2.2.10-canary.15.
CVE-2025-71385 is a reflected cross-site scripting (XSS) vulnerability in netdata versions before 2.3.1. The vulnerability occurs because the user-supplied 'love' query parameter in the /api/v2/ilove.svg and /api/v3/ilove.svg endpoints is included verbatim in the generated SVG document without proper escaping. These endpoints are accessible without authentication by default, allowing attackers to craft malicious URLs that execute scripts in the victim's browser within the netdata origin. The issue was fixed by removing the vulnerable ilove endpoints in version 2.3.1.
Crawl4AI versions before 0.8.7 have a critical vulnerability in the Docker API server's /execute_js endpoint that allows arbitrary JavaScript execution. This endpoint executes user-supplied JavaScript in the server's browser context with disabled web security, enabling attackers to perform server-side request forgery against internal services.
Invoice Ninja versions up to 5.13.26 contain an open redirect vulnerability in the client portal login. This flaw allows unauthenticated attackers to craft login links with malicious external URLs in the intended parameter. After a victim logs in legitimately, they are redirected to the attacker-controlled URL without validation, enabling phishing attacks.
Presenton versions before 0.8.8-beta include a bundled MCP server that is accessible without authentication on the /mcp path when deployed with session authentication via nginx. This occurs because the nginx front-end does not enforce the auth_request gate on /mcp, allowing remote unauthenticated attackers to invoke MCP tools and perform actions such as generating presentations and consuming configured LLM API keys. The Electron desktop build is not affected as MCP is disabled there. A fix is available in version 0.8.8-beta.
JeecgBoot versions up to 3.9.2 have a broken access control vulnerability that allows authenticated low-privilege users to fully manage OpenAPI credentials. The affected endpoints lack proper authorization checks, exposing secret keys in plaintext and enabling unauthorized credential theft and API invocation.
CVE-2026-58376 is a SQL injection vulnerability in Dolibarr versions up to 23.0.3. It affects authenticated API users who can exploit the sqlfilters query parameter in the setup dictionary and multicurrencies REST API endpoints to exfiltrate arbitrary database contents. The flaw arises because the endpoints validate sqlfilters only for balanced parentheses and rewrite matched triplets, but allow appended SQL commands such as UNION SELECT to be concatenated unmodified into the SQL WHERE clause. This enables attackers to retrieve sensitive data including password hashes and API keys. The vulnerability is rated high severity with a CVSS score of 7.6. A fix was introduced in commit 14db36e after version 23.0.3.
SeaweedFS versions prior to 4.34 contain a path traversal vulnerability in the S3 gateway's DeleteMultipleObjectsHandler. Authenticated users with write access to a single bucket can exploit this flaw by including '../' sequences in object keys within the DeleteObjects XML request body. This allows deletion of objects outside their authorized bucket, bypassing authorization controls due to insufficient validation of request body keys. The vulnerability is fixed in version 4.34.
CVE-2026-58371 is a cross-site scripting (XSS) vulnerability in SeaweedFS versions before 4.30. It arises from improper neutralization of the callback query parameter in JSON endpoints, which are served with Content-Type application/javascript without validation or security headers. This allows a third-party web page to load these endpoints cross-origin via a script tag and read sensitive cluster information such as topology, volume server URLs, gRPC ports, file identifiers, and directory listings. The vulnerability affects multiple unauthenticated endpoints accessible by default configurations. The issue is fixed in SeaweedFS version 4.30.
CVE-2026-57997 is a vulnerability in the Strapi users-permissions plugin where the JWT algorithm restrictions are not enforced unless explicitly configured. This allows acceptance of HS384 and HS512 tokens in addition to HS256. Attackers with knowledge of the jwtSecret can create tokens using these non-standard HMAC algorithms to bypass authentication controls. The issue affects Strapi versions prior to 5.7.0 and has been fixed in version 5.7.0.
Papermark versions up to 0.22.0 have a CORS misconfiguration vulnerability allowing unauthenticated attackers to perform credentialed cross-origin requests via the TUS-based viewer upload endpoint. This misconfiguration reflects arbitrary Origins with Access-Control-Allow-Credentials set to true, enabling attackers to trick authenticated users into uploading arbitrary files to victim datarooms and reading credentialed responses.
CVE-2026-57956 is an authorization bypass vulnerability in SigNoz versions before 0.133.0. It allows authenticated users to access and manipulate alert rules belonging to other organizations by exploiting missing tenant isolation checks. This flaw enables reading, editing, and deleting alert rules across organizational boundaries, bypassing multi-tenant access controls.
PhotoPrism versions before 260601-a7d098548 contain an access control vulnerability allowing authenticated non-admin users to modify other users' profile information. This occurs due to missing session-to-user identifier validation in the PUT users API endpoint, enabling unauthorized profile overwrites.
LibrePhotos versions before 1.0.0 have a broken object-level authorization vulnerability in the SetPhotosShared endpoint. This flaw allows authenticated users to bypass ownership validation and grant themselves access to other users' private photos by manipulating shared_to relations. The vulnerability is fixed in version 1.0.0.