CVE Database

Incorrect permission assignment for critical resource in M365 Copilot allows an authorized attacker to disclose information over a network.

Improper neutralization of input during web page generation ('cross-site scripting') in Azure Portal allows an unauthorized attacker to perform spoofing over a network.

vLLM through 0.29.0 fails to properly clean up decode-side metadata for rejected inference requests in prefill/decode disaggregated deployments. Remote attackers can submit requests with max_tokens=0 to exhaust decode-worker memory without bound until the worker restarts.

redis-parser through 3.0.0 contains a denial of service vulnerability in the RESP protocol parser that allows malicious Redis endpoints to crash the client process through unbounded recursion on nested arrays. Attackers can send crafted RESP byte streams with repeated array headers that exhaust the V8 call stack, causing an uncaught RangeError that terminates the Node.js process without triggering error handling callbacks.

Local privilege escalation due to insecure file permissions. The following products are affected: Acronis Backup plugin for cPanel & WHM (Linux) before build 1.9.3.1021, Acronis Backup extension for Plesk (Linux) before build 1.8.11.638, Acronis Backup plugin for DirectAdmin (Linux) before build 1.2.3.238.

Server-side request forgery (ssrf) in Azure AI Foundry allows an unauthorized attacker to elevate privileges over a network.

Improper neutralization of special elements used in a command ('command injection') in Microsoft 365 Copilot's Business Chat allows an unauthorized attacker to disclose information over a network.

Incorrect authorization in Azure Machine Learning allows an unauthorized attacker to disclose information over a network.

SigNoz versions 0.87.0 before 0.142.0 fail to escape user-supplied telemetry field-key names in the v5 query_range API, allowing authenticated users to inject SQL. Attackers with Viewer role or higher can embed backticks and quotes in field names to break out of identifiers and string literals, executing arbitrary ClickHouse SQL to read system tables and exfiltrate data.

Session Fixation vulnerability in team-alembic ash_authentication allows an attacker who can plant a session identifier in a victim's browser to hold an authenticated session once that victim signs in. AshAuthentication.Plug.Helpers.store_in_session/2 writes the authenticated subject into the existing session with Plug.Conn.put_session/3 and never calls Plug.Conn.configure_session(renew: true), so the identifier the visitor arrived with carries into their authenticated session. Every authentication event reaches this one function: the default success/4 injected by AshAuthentication.Phoenix.Controller.__using__/1, the AuthController emitted by mix ash_authentication_phoenix.install, and remember-me auto-login. AshAuthentication.Phoenix.Plug.store_in_session/2 is a defdelegate to it. Logout does not close the window either, because clear_session/2 ends with Plug.Conn.clear_session/1, which clears session contents but leaves the identifier intact, so a planted identifier survives a logout-then-login cycle. This issue affects ash_authentication: from 0.2.0 before 4.15.0 and from 5.0.0-rc.0 before 5.0.0-rc.14.

WeGIA is a web manager for charitable institutions. Prior to 3.8.5, WeGIA maps InternoControle to an empty resource array in web/controle/control.php, and verificarPermissao in web/dao/MiddlewareDAO.php treats that empty array as unconditional access for every authenticated user. The methods in web/controle/InternoControle.php, including listarUm, alterar, and excluir, accept user-controlled id or idInterno values without verifying ownership, allowing a low-privileged user to read, modify, or delete another person's records and expose personal, identity, address, medical, and family information. The advisory notes that a self-referencing load bug can crash this controller in the reported revision, but the empty-resource authorization pattern and affected methods remain the vulnerability under review. This issue is fixed in version 3.8.5.

CubeCart is an ecommerce software solution. Prior to 6.7.5, admin/sources/settings.index.inc.php directly concatenates the administrator-controlled download_expire POST parameter into a raw UPDATE statement for CubeCart_downloads without numeric validation. An authenticated administrator can supply a comma-delimited value that changes the SET clause because HTML sanitization does not neutralize SQL syntax, allowing manipulation of database columns and potentially other data within the application's database privileges. This issue is fixed in version 6.7.5.

CubeCart is an ecommerce software solution. Prior to 6.7.5, admin/sources/maintenance.index.inc.php places administrator-controlled tablename values into ALTER TABLE, CHECK TABLE, and ANALYZE TABLE statements without validating the identifiers or escaping embedded backticks. An authenticated administrator can terminate the quoted identifier with a closing backtick and introduce attacker-controlled structural SQL, potentially compromising database confidentiality, integrity, and availability within the application's database privileges. This issue is fixed in version 6.7.5.

Hamlib is a ham radio control library for radios, rotators, and amplifiers. Prior to 4.7.2, the unauthenticated rigctld send_raw command on TCP port 4532 reaches rigctl_send_raw() in tests/rigctl_parse.c, which writes a NUL byte at buf[buf_len + 1] outside its 200-byte stack buffer, and rig_send_raw() in src/rig.c, which copies reply_len - 1 bytes instead of the actual nbytes received. A remote client can send the CR terminator with a short payload to trigger both flaws in one command under the default no-password configuration. The out-of-bounds write can crash the daemon or corrupt adjacent stack memory, while the oversized copy can return up to 198 bytes of uninitialized stack data to the client. This issue is fixed in version 4.7.2.

Vvveb is a powerful and easy to use CMS with page builder to build websites, blogs or ecommerce stores. From 1.0.0 until 1.0.8.5, saveGlobalElements() in admin/controller/editor/global-trait.php concatenates the attacker-controlled file portion of data-v-save-global to the active theme directory before loadHTMLFile() and file_put_contents() operate on it. An authenticated user with the default Editor role and editor/* permission can submit crafted HTML to module=editor/editor&action=save and traverse to an existing writable PHP file outside the theme directory. If the target is web-accessible, editor-controlled PHP content executes in the web server context; a shipped public/vadmin/index.php entrypoint can be used as an execution trampoline rather than requiring a test-only file. This can permit persistent webshell placement and compromise application confidentiality, integrity, and availability. This issue is fixed in version 1.0.8.5.

MythicalDash is a Pterodactyl client area. In 3.5.4-aurora and earlier, GET /api/stripe/process in backend/app/Api/System/Gateways/Stripe.php creates a pending row in mythicaldash_stripe_payments before Stripe checkout succeeds and embeds the payment code in the success redirect, while GET /api/stripe/processed accepts that code without constructing a Session, checking ownership, or retrieving the Stripe Checkout Session to require payment_status to be paid and amount_total to match the expected charge. An ordinary authenticated user can request an attacker-selected coins amount, abandon or fail payment, and submit the pending code directly to the unauthenticated processed endpoint. StripeDB::isPending() then permits User::addCreditsAtomic() to grant the unpaid amount and mark the row processed even though Stripe has not confirmed payment. This permits arbitrary free virtual-currency top-ups and direct financial loss through consumption of hosting resources. No fixed version is available as of this review.

AI Agent Automation is a modular AI agent workflow automation platform with schedulers, tools, and observability. Prior to 0.9.1, the executeStep file-step implementation in backend/src/agents/executor.js passes the user-controlled step.path value through path.resolve with process.cwd() and then uses the resulting path for read or write operations without checking that it remains in an approved workflow directory. An authenticated user who can create or modify workflow file steps can supply traversal segments to escape the intended workspace and read sensitive files or write and overwrite files accessible to the backend process, including application-adjacent files when process permissions allow. This issue is fixed in version 0.9.1.

AI Agent Automation is a modular AI agent workflow automation platform with schedulers, tools, and observability. Prior to 0.9.1, backend/src/controllers/memory.controller.js authenticates requests but listMemories, deleteMemory, and clearAgentMemory use a caller-supplied agentId or memory _id without verifying through the related Agent that the record belongs to req.user. An authenticated attacker who knows or obtains another user's identifiers can read victim AgentMemory content, including conversation history, agent context, task data, embeddings, and metadata, delete an individual victim memory, or clear all memory belonging to a victim agent. This breaks tenant isolation and causes unauthorized disclosure and data loss. This issue is fixed in version 0.9.1.

Vvveb is a powerful and easy to use CMS with page builder to build websites, blogs or ecommerce stores. Prior to 1.0.8.5, the oEmbedProxy() handler in admin/controller/editor/editor.php accepts an attacker-controlled url parameter and passes it to getUrl(), while validateUrl() in system/functions.php checks only the hostname string and does not validate its resolved addresses. An authenticated admin-panel user with editor/* permission can invoke GET /admin/index.php?module=editor/editor&action=oEmbedProxy with a dotted hostname or normalized loopback form that resolves to a private, loopback, link-local, or reserved address, causing the server to issue an HTTP or HTTPS request and return the response body. Storefront users and anonymous visitors cannot invoke the endpoint, but no CSRF token is required because the action uses GET. This can disclose internal service responses or cloud instance metadata and associated credentials. This issue is fixed in version 1.0.8.5.

Vvveb is a powerful and easy to use CMS with page builder to build websites, blogs or ecommerce stores. Prior to 1.0.8.5, app/controller/user/profile.php accepts the user[bio] field and passes stored content through sanitizeHTML() in system/functions.php, whose on* event-handler regular expression omits the forward-slash delimiter and whose do-while condition compares the string to itself, so forbidden nested tags are removed only once. An Author-role or higher user can submit solidus-prefixed event-handler markup or nested forbidden tags that survive sanitization. The stored bio is rendered without sufficient output encoding on /author/{username}, in the admin user-management view, and potentially in comment displays, causing attacker-controlled JavaScript to execute when unauthenticated visitors, administrators, or other users view the content. This can expose browser-session data and permit victim-context account actions, defacement, or phishing. This issue is fixed in version 1.0.8.5.

Frappe Learning Management System (LMS) is a learning system that helps users structure their content. Prior to version 2.52.1, a remote attacker can request a traversal path handled by SCORMRenderer.render in lms/page_renderers.py. The renderer constructs and opens a server-side path without first confirming that its real path remains within public/scorm, allowing files outside the SCORM directory to be read when they are accessible to the server process. This issue is fixed in version 2.52.1.

SQLBot is an intelligent Text-to-SQL system based on large language models and RAG. Prior to 1.9.0, an authenticated user can supply a crafted sheet["tableName"] value in the Excel datasource configuration submitted through POST /api/v1/datasource/, and SQLBot stores that value without safe identifier handling. When the same datasource is later removed through DELETE /api/v1/datasource/{id}, the stored value is interpolated into datasource cleanup SQL and executed by PostgreSQL. This second-order SQL injection can invoke PostgreSQL COPY TO PROGRAM and execute arbitrary operating-system commands with the privileges of the postgres process inside the SQLBot container. This issue is fixed in version 1.9.0.

SQLBot is an intelligent Text-to-SQL system based on large language models and RAG. Prior to 1.9.0, the POST /api/v1/datasource/parseExcel endpoint in backend/apps/datasource/api/datasource.py uses attacker-controlled multipart filename data when selecting where uploaded content is stored, writes the content before spreadsheet parsing and validation finish, and can transform a double-extension filename into a Python source file. An attacker able to submit a crafted multipart upload can use these behaviors to place attacker-controlled content in /opt/sqlbot/app/alembic/versions/ even when a spreadsheet parsing failure after the file write causes the endpoint to return an error. The planted file remains on disk, and subsequent SQLBot startup or migration processing causes Alembic to import the module and execute its module-level statements in the SQLBot application runtime. This issue is fixed in version 1.9.0.

JabRef is a desktop application for managing BibTeX and BibLaTeX libraries. Prior to 6.0-alpha.6, when jabsrv or JabRef's built-in HTTP server is enabled, the GET /better-bibtex/cayw endpoint accepts an external command query parameter and CAYWQueryParams.getCommand() passes it through CAYWResource.getCitation() into PushToSublimeText.getCommandLine(). On Unix-like systems, PushToSublimeText combines this untrusted cite-command prefix and citation keys into a string executed through sh -c by ProcessBuilder without shell escaping. A client that can cause a localhost request with application=sublime can inject shell metacharacters and execute operating-system commands as the JabRef user when a valid Sublime Text command path is configured and the victim completes the CAYW selection dialog. The built-in server is disabled by default, so exploitation requires the victim to enable it or run jabsrv. This issue is fixed in version 6.0-alpha.6.

yutu is an AI-powered toolkit for managing and growing YouTube channels. Prior to 0.10.9, the caption-download MCP tool accepts a caller-controlled file parameter through cmd/caption/download.go and passes it to Caption.Download() in pkg/caption/caption.go, where os.Create() creates or truncates that path without using the pkg.Root confinement boundary backed by YUTU_ROOT. A principal able to invoke caption-download, including a local HTTP client when the MCP server runs with its default authentication-disabled configuration, can write downloaded caption bytes to any path writable by the yutu process outside YUTU_ROOT. This can overwrite application files, configuration, shell startup files, logs, or data and can cause persistent code execution or denial of service depending on the selected writable target. Live caption retrieval also requires usable service credentials and an accessible caption identifier. This issue is fixed in version 0.10.9.

A heap-based buffer overflow exists in the TLS transport layer of the MongoDB C Driver when built with the Windows platform TLS backend. A remote endpoint that the client connects to can cause the driver to write uncontrolled data outside the bounds of a heap allocation while processing incoming encrypted traffic after the TLS handshake completes. No authentication or user interaction is required, because the affected processing occurs before any application-level authentication completes. Triggering this issue may lead to memory corruption in the client process, disclosure of adjacent heap memory, or termination of the process.

Use after free in PDFium in Google Chrome prior to 153.0.8010.52 allowed a remote attacker to execute arbitrary code inside the sandbox via a crafted HTML page. (Chromium security severity: High)

Buffer overflow in PDFium in Google Chrome on on Windows prior to 153.0.8010.52 allowed a remote attacker leveraging social engineering to potentially execute arbitrary code inside the sandbox via a crafted PDF file. (Chromium security severity: High)

Type confusion in V8 in Google Chrome prior to 153.0.8010.52 allowed a remote attacker leveraging social engineering to execute arbitrary code inside the sandbox via a crafted HTML page. (Chromium security severity: High)

Incorrect reference resolution in Tracing in Google Chrome on on Windows prior to 153.0.8010.52 allowed a local attacker to potentially execute arbitrary code outside the sandbox via a local program. (Chromium security severity: High)