Search vulnerabilities

Results 5,758

Export CSV
CVE-2026-54031

LibreChat: MCP OAuth requests can follow redirects to internal URLs after SSRF validation

LibreChat is an enhanced ChatGPT clone that supports multiple AI providers. Prior to 0.8.8, packages/api/src/mcp/oauth/handler.ts createOAuthFetch validates only the initial destination for MCP OAuth discovery, registration, token, refresh, and revocation requests, then uses redirect-following fetch behavior without validating each redirect target. An attacker who can initiate MCP OAuth against a controlled endpoint can redirect server requests to internal or private addresses, reaching services that the initial SSRF checks should block. The separate MCP transport path already validates redirects and is not affected. This issue is fixed in version 0.8.8.

CVSS 3.1 · 6.3 Medium EPSS < 1% KEV · Not listed
Vendors
—
Products
—
CVE-2026-54039

LibreChat: Pre-registration account takeover via attacker-controlled `provider` field and social-login email fallback

LibreChat is an enhanced ChatGPT clone that supports multiple AI providers. From 0.7.0 until 0.8.6, when local registration is enabled and at least one social provider is configured, POST /api/auth/register passes an unparsed request object into api/server/services/AuthService.js, allowing an unauthenticated attacker to persist a forged provider value for a known victim email. The api/strategies/socialLogin.js email-fallback path then accepts the planted record when the victim signs in because it compares the stored provider name without requiring the provider subject identifier. The victim OAuth session and the attacker local-password session become bound to the same user record, giving the attacker persistent read and write access to data created by the victim. The first-user administrator role is an additional conditional impact only on a fresh deployment. This issue is fixed in version 0.8.6.

CVSS 3.1 · 8.1 High EPSS < 1% KEV · Not listed
Vendors
—
Products
—
CVE-2026-95825

Reference-count underflow in Zephyr network interface address removal releases an IP address that is still in use

net_if_ipv4_addr_rm() and net_if_ipv6_addr_rm() in subsys/net/ip/net_if.c unconditionally called net_if_addr_unref() on the target address. When an earlier removal could not complete because another subsystem still held a reference (the TCP stack takes one on the local address of every connection, in tcp_conn_new() and on connect()), the address is left in place with is_added cleared and is_used still set. A second removal of such an address decremented the reference count again, consuming a reference owned by that other holder. When it was the last one, the address slot was released, its NET_EVENT_IPV*_ADDR_DEL event fired, and the slot became reusable while the holder still believed it owned a reference. Every address lookup helper matches on is_used and not on is_added, so a half-removed address stays visible to the rest of the stack and repeat removals are readily reachable from remote input. Under IPv6 SLAAC, an on-link node that sends a Router Advertisement with a short valid lifetime causes the address to expire into the deferred state; a refreshing Router Advertisement is handled by net_if_ipv6_addr_update_lifetime(), which re-arms the lifetime timer without restoring the reference, so the next expiry performs the second, reference-stealing removal. With CONFIG_NET_IPV4_ACD and DHCPv4, repeated spoofed ARP conflicts for the device's lease drive dhcpv4_acd_event_handler() into calling net_if_ipv4_addr_rm() more than once on the same address. Where CONFIG_USERSPACE is enabled, the net_if_ipv4_addr_rm_by_index() and net_if_ipv6_addr_rm_by_index() system calls let an unprivileged user-mode thread trigger the same kernel-side release deterministically. The impact is loss of a configured IP address while connections are still using it, and a corrupted reference count that propagates: the slot can be handed to a newly added address, and when the stale holder later tears down, net_if_addr_unref() resolves the address by value and drops the new incarnation's only reference, removing a live address a second time. The affected objects are static array slots and the counter is never decremented below zero, so there is no memory corruption or information disclosure; the consequence is denial of service against the device's network connectivity, repeatable at the attacker's pace. The fix resolves the address before touching the reference count and returns false without calling net_if_addr_unref() when the address is already marked as removed, so the public removal API can no longer consume a reference it does not own.

CVSS 3.1 · 5.3 Medium EPSS < 1% KEV · Not listed
Vendors
Products
CVE-2026-55372

NukeViet CMS: Pre-authentication SSRF via X-Forwarded-Host in NukeViet

NukeViet CMS is a multi Content Management System. Prior to 4.5.09, an unauthenticated attacker can submit the __serverInfoUpdate=1 field while controlling X-Forwarded-Host and X-Forwarded-Proto. NukeViet\Core\Server::standardizeHost() did not validate the forwarded host against my_domains, and a trailing slash could preserve an attacker-selected port that server_info_update() used to construct a cURL request in includes/ini.php before authentication. The blind, HEAD-only request uses a fixed path, but it permits internal host and port discovery and can poison cached server_headers. This issue is fixed in version 4.5.09.

CVSS 3.1 · 7.2 High EPSS < 1% KEV · Not listed
Vendors
Products
CVE-2026-133875

one-hub through 0.14.27 SSRF via Vision Chat image_url

one-hub through 0.14.27 contains a server-side request forgery vulnerability in GetImageFromUrl in common/image/image.go that fetches vision chat image_url values without destination restrictions. Attackers holding a standard non-admin API token can send crafted /v1/chat/completions requests via Claude, OpenRouter or Ollama providers to reach internal services or cloud metadata endpoints.

CVSS 4.0 · 5.3 Medium EPSS < 1% KEV · Not listed
Vendors
Products
CVE-2026-129029

WordPress OttoKit plugin <= 1.1.42 - Server Side Request Forgery (SSRF) vulnerability

Server-Side Request Forgery (SSRF) vulnerability in Brainstorm Force OttoKit suretriggers allows Server Side Request Forgery.This issue affects OttoKit: from n/a through 1.1.42.

CVSS 3.1 · 7.2 High EPSS < 1% KEV · Not listed
Vendors
—
Products
—
CVE-2026-104184

LinkAce: Mass Assignment via user_id in API Link Update Allows Ownership Transfer and Content Injection

LinkAce is a self-hosted archive to collect website links. Prior to 2.6.0, LinkController::update() in app/Http/Controllers/API/LinkController.php passes $request->all() to the repository update method even though user_id is fillable and is not an allowed field in the validated update request. An authenticated API user who is authorized to update a link can therefore assign an arbitrary or nonexistent user_id, transferring the link into another user's bookmark collection or orphaning it from any real account. The web update controller is unaffected because it passes only validated fields. This issue is fixed in version 2.6.0.

CVSS 3.1 · 7.1 High EPSS < 1% KEV · Not listed
Vendors
Products
CVE-2026-133706

request-filtering-agent before 3.2.2 SSRF Filter Bypass via IPv4-Compatible IPv6 Addresses

request-filtering-agent before 3.2.2 contains a server-side request forgery filter bypass because address validation in src/request-filtering-agent.ts treats IPv4-compatible IPv6 addresses in the ::/96 range as public unicast. Attackers controlling a fetched URL can use addresses like ::7f00:1 or ::a9fe:a9fe to reach loopback or link-local services, including cloud metadata endpoints.

CVSS 4.0 · 6.3 Medium EPSS < 1% KEV · Not listed
Vendors
Azu
CVE-2026-77380

Bolt: Prevent SSRF due to unchecked followed redirects

Bolt is a fully open source, adaptable platform for building and running modern websites. Prior to 6.1.5, the /bolt/async/upload-url and /bolt/async/embed endpoints validate only a submitted URL before fetching it server-side. The fetchers then follow redirects without revalidating each destination, allowing an authenticated user to bypass UrlSafetyChecker and cause the server to request internal or cloud-metadata addresses. This issue is fixed in version 6.1.5.

CVSS 3.1 · 5.2 Medium EPSS < 1% KEV · Not listed
Vendors
Products
CVE-2026-59988

django-docutils unsafe Docutils defaults allow local file disclosure, SSRF, and stored XSS

django-docutils provides reStructuredText support for Django. Prior to 0.30.0, public rendering paths including publish_doctree, publish_html_from_source, publish_html_from_doctree, publish_parts_from_doctree, the {% rst %} template tag, the rst template filter, DocutilsTemplates, and DocutilsView pass lower-trust reStructuredText to Docutils without disabling file_insertion_enabled, raw_enabled, ambient docutils.conf processing, or unsafe URI schemes and without consistently sanitizing doctrees. An attacker who can influence rendered reStructuredText can use include or csv-table directives to read local files or cause server-side URL requests, or use raw HTML and unsafe URI output to inject stored script-capable content. Applications that render only trusted static reStructuredText are not affected unless attackers can influence it. This issue is fixed in version 0.30.0.

CVSS 3.1 · 8.2 High EPSS < 1% KEV · Not listed
Vendors
Products
CVE-2026-133276

Validation bypass of Pydantic V1 dataclasses allowing extra data

When extra data is allowed (extra=Extra.allow), providing a __pydantic_initialised__ key when validating a Pydantic V1 dataclass makes it considered as already validated, resulting in field values not being validated. The __init__() method of Pydantic V1 dataclasses stores all the provided keyword arguments in the instance __dict__ before validation, so a __pydantic_initialised__ keyword argument overrides the internal flag used to skip validation of already validated instances. As a result, data returned by validation doesn't match the declared type. Affected are applications validating untrusted data into Pydantic V1 dataclasses allowing extra data. The issue was fixed in Pydantic 1.10.27. Users relying on the pydantic.v1 namespace in Pydantic V2 should upgrade to 2.14.2 or later.

CVSS 4.0 · 6.3 Medium EPSS < 1% KEV · Not listed
Vendors
Products
CVE-2026-133237

Manyfold 0.124.0 before 0.154.2 SSRF via ActivityPub Actor Icon and Image URLs

Manyfold 0.124.0 before 0.154.2 contains a server-side request forgery vulnerability that allows unauthenticated remote attackers to make the instance fetch internal URLs by supplying malicious ActivityPub actor icon.url and image.url fields. Attackers can deliver Create or Update activities to the inbox, causing the Shrine remote_url plugin to request internal or cloud metadata hosts and store responses as creator avatar or banner attachments.

CVSS 4.0 · 6.9 Medium EPSS < 1% KEV · Not listed
Vendors
Products
CVE-2026-133235

Tandoor Recipes before 2.7.4 SSRF via Nextcloud Storage WebDAV Client

Tandoor Recipes before 2.7.4 contains a server-side request forgery vulnerability that allows space admins to reach internal hosts because Nextcloud.get_client builds WebDAV clients from Storage.url without calling safe_request. Attackers can create a NEXTCLOUD storage via POST /api/storage/ pointing at internal hosts, import a file, then call GET /api/get_recipe_file/<id>/ to read internal service or cloud metadata responses.

CVSS 4.0 · 8.3 High EPSS < 1% KEV · Not listed
Vendors
Products
CVE-2026-92804

Nango through 0.72.6 Server-Side Request Forgery via Configuration

Nango through 0.72.6 fails to validate caller-supplied connection configuration values interpolated into provider token and proxy URL templates. Authenticated attackers can supply malicious configuration values to direct server requests at internal addresses or cloud metadata endpoints (0.71.3 and later block cloud metadata and loopback addresses by default but still permit private network addresses), potentially exfiltrating provider credentials.

CVSS 4.0 · 7.1 High EPSS < 1% KEV · Not listed
Vendors
Products
CVE-2026-86539

knowns through 0.36.1 Server-Side Request Forgery via embedding-models endpoint

knowns through 0.36.1 contains a server-side request forgery vulnerability in the POST /api/embedding-models/test endpoint that issues outbound requests to caller-supplied destinations without validation. Attackers can enumerate internal hosts and cloud metadata endpoints by observing transport error messages that reveal network reachability information.

CVSS 4.0 · 6.9 Medium EPSS < 1% KEV · Not listed
Vendors
Products
CVE-2026-81091

mcp-use Inspector Proxy Server-Side Request Forgery via Caller-Supplied Target URL

The proxy middleware in mcp-use's inspector forwards requests to a destination the caller names. mountMcpProxy in libraries/typescript/packages/mcp-use/src/server/middleware/mcp-proxy.ts read the target from the X-Target-URL header or the __mcp_target parameter and proxied to it without inspecting the host, so loopback, link-local and private addresses were all accepted, as were names that resolve to them, and the validation was not reapplied to a redirect the destination returned. A caller could therefore make the server issue requests to addresses reachable only from the host it runs on and read the responses. In mcp-use 2.x the proxy moved to the @mcp-use/inspector package, whose mcp-proxy.ts calls isSafeProxyTarget, which checks the resolved address against private, loopback and link-local ranges before proxying and bounds the number of redirects followed.

CVSS 4.0 · 8.7 High EPSS < 1% KEV · Not listed
Vendors
Products
CVE-2026-39922

GeoNode SSRF via Service Registration

GeoNode versions 4.4.5 and 5.0.2 (and prior within their respective releases) contain a server-side request forgery vulnerability in the service registration endpoint that allows authenticated attackers to trigger outbound network requests to arbitrary URLs by submitting a crafted service URL during form validation. Attackers can probe internal network targets including loopback addresses, RFC1918 private IP ranges, link-local addresses, and cloud metadata services by exploiting insufficient URL validation in the WMS service handler without private IP filtering or allowlist enforcement.

CVSS 4.0 · 5.3 Medium EPSS < 1% KEV · Not listed
Products
CVE-2026-110628

Kan 0.5.5 through 0.7.0 Unauthenticated SSRF via /api/download/attatchment

Kan 0.5.5 through 0.7.0 contains an unauthenticated server-side request forgery vulnerability in the /api/download/attatchment route due to an incomplete fix for CVE-2026-32255. When S3_ENDPOINT is unset, attackers can pass arbitrary url query parameters to fetch() to read internal services, localhost ports, and cloud instance metadata.

CVSS 4.0 · 9.2 Critical EPSS < 1% KEV · Not listed
Vendors
Kan
Products
Kan
CVE-2026-101883

OpenClaw Windows Node through 2026.11.3 SSRF via canvas.present

OpenClaw Windows Node through 2026.11.3 contains a server-side request forgery vulnerability in the canvas.present capability that bypasses URL risk evaluation enforced by canvas.navigate. Attackers with gateway or agent access can issue canvas.present to make the node's WebView send requests to localhost, private networks, or tailnet services from the user's machine.

CVSS 4.0 · 5.3 Medium EPSS < 1% KEV · Not listed
Vendors
CVE-2026-99322

ksmbd: rebind completed CREATE replays without draining the old channel

In the Linux kernel, the following vulnerability has been resolved: ksmbd: rebind completed CREATE replays without draining the old channel A completed DurableHandleReqV2 CREATE can be replayed through another channel while its original channel is exiting. The existing replay path returns the open but leaves its connection and oplock references pointing at the old channel. Its open count also stays on that channel, and a later CLOSE through the surviving channel decrements the wrong counter. After checking the owner, lease key and persistent flag, verify that an active open is still published in the requesting session's file table. Do this even when no oplock was granted, and before processing a consumed replay. Do not read a borrowed opinfo session pointer for this check. Keep a healthy original binding. If the original channel is exiting or releasing, move the open and its own oplock references to the requesting channel. Serialize this with CLOSE using the session file-table lock, and with detach, reconnect and notification snapshots using m_lock. The earlier lock index and lease-table changes make their lifetimes independent of this connection change. No old-channel r_count wait is needed, including when an unrelated request remains active there. Account final close against fp->conn in one place, balance durable detach and reconnect, and transfer the count with a replay binding change. A final global lookup put may run after CLOSE has removed the volatile id; unlink the open from the inode list on that final close as well. A live completed CREATE replay may follow a break that lowered the durable open's batch oplock. Apply the batch eligibility check only to detached reconnects, while retaining the owner, lease key and persistent flag checks for active replays.

CVSS 3.1 · 6.5 Medium EPSS < 1% KEV · Not listed
Vendors
Products