| CVE |
Vendors |
Products |
Updated |
CVSS v3.1 |
| Headroom's LLM proxy lets a client choose the upstream destination with the x-headroom-base-url request header. _resolve_openai_upstream_base in headroom/proxy/handlers/openai.py accepts the header value, requires only that it parse with an http or https scheme and a hostname, and returns it for use as the upstream base; _select_passthrough_base_url in headroom/providers/proxy_routes.py reads the same header for the passthrough routes. No check rejects loopback, link-local, or RFC 1918 destinations, and because the component is a proxy the upstream response is returned to the caller, so the request reaches internal services and cloud metadata addresses and their responses are disclosed. The Authorization header accompanying the request is forwarded unchanged to the caller-designated host. The pip console script binds 127.0.0.1 by default, but the reference docker-compose.yml ships --host 0.0.0.0 with published ports and no required HEADROOM_PROXY_TOKEN, which the server itself warns about at startup, so a deployment following the shipped compose exposes the affected data-plane routes to the network without authentication. |
| n8n before 1.123.69, 2.33.4, and 2.34.1 contains an SSRF protection bypass in the OAuth2 credential authorization-code-to-access-token exchange. While OAuth2 discovery and dynamic-client-registration requests use n8n's SSRF-protected HTTP client, the token exchange uses a separate client with no SSRF guard. A user with credential-creation permissions can set the access-token URL to an internal address and complete the OAuth2 flow, causing n8n to send a fixed-shape token-exchange POST to that target and reflect its response body back to the attacker (limited to what the target returns to this specific request). |
| Server-side request forgery (ssrf) in Microsoft Exchange Online allows an unauthorized attacker to elevate privileges over a network. |
| The setWebhookResolver in packages/api/src/resolvers/webhooks/index.ts stores the caller-supplied url without any address validation, and the file imports no validation helper. When a subscribed event fires, callWebhook in packages/api/src/jobs/call_webhook.ts issues axios.request with that url, the method and Content-Type recorded on the webhook, and a JSON body carrying the event data, so an authenticated user can make the server send repeated attacker-shaped requests to internal endpoints, including link-local metadata addresses. The request is blind: callWebhook discards the result and writes only a success line or the axios error to the server log, so the response is not returned through the API. |
| Authenticated pre-validation SSRF vulnerability in Apache CloudStack's template and ISO registration functionality.
When registering a template or ISO, CloudStack makes a live HTTP HEAD/GET call to determine file size for secondary storage usage-limit checks, and this happens before URL validation is performed. However, this does not pose a malicious template or ISO registration risk, as URL validation still occurs prior to the actual download by the Secondary Storage VM.This issue affects Apache CloudStack: in 4.20.3.0 and from 4.21.0.0 through 4.22.1.0.
Users are recommended to upgrade to version 4.20.3.1 or 4.22.1.1 or later, which fixes the issue. |
| Server-Side Request Forgery (SSRF) vulnerability in Apache CloudStack's webhook module, exploitable via webhook delivery requests.
This issue affects Apache CloudStack: from 4.20.0.0 through 4.20.3.0 and from 4.21.0.0 through 4.22.1.0.
Users are recommended to upgrade to version 4.20.3.1 or 4.22.1.1 or later, which fixes the issue. |
| TREK is a collaborative travel planner. Prior to 3.1.0, TREK validates only the initial URL before native redirect following in importGoogleList() and importNaverList() in server/src/services/placeService.ts and resolveGoogleMapsUrl() in server/src/services/mapsService.ts. The affected sinks call checkSsrf() from server/src/utils/ssrfGuard.ts and then use fetch() with redirect: 'follow' instead of the DNS-pinned safeFetch() path, so a public attacker-controlled URL can redirect the server to loopback, RFC 1918, or cloud metadata addresses without revalidation. An authenticated trip member can reach the list-import routes, and any authenticated user can reach /api/maps/resolve-url, allowing blind GET requests to internal services without response-body reflection. This issue is fixed in version 3.1.0. |
| In OpenStack Glance through 32.0.0, the /v2/tasks API accepts type=import tasks that
bypass import_filtering_opts, allowing an admin to fetch internal
URLs from the Glance service network (aka SSRF), as long as https:// or http:// is used. This API has been available only to admins since Xena, and it has been deprecated for several releases. |
| The POST /api/provider-nodes/validate route in 9router takes a caller-supplied baseUrl and issues server-side HTTP requests to it, guarding the destination with assertPublicUrl from src/shared/utils/ssrfGuard.js. That guard compares hostname strings only: it resolves no DNS, does not revalidate after a redirect, and its IPv4-mapped IPv6 branch is unreachable. The branch matches ^::ffff:(\d+\.\d+\.\d+\.\d+)$, but the WHATWG URL parser canonicalizes such literals to hextets before the guard runs, so new URL("http://[::ffff:127.0.0.1]/").hostname yields [::ffff:7f00:1] and the pattern is tested against a string it is never handed. Every IPv4-mapped address therefore passes, and http://[::ffff:7f00:1] and http://[::ffff:a9fe:a9fe] reach loopback and link-local metadata addresses; a hostname whose A record points at an internal address passes as well because no resolution occurs. In the custom-embedding branch the upstream response body is truncated to 200 bytes and returned to the caller whenever the upstream status is neither 2xx nor 401 nor 403, which discloses the beginning of internal responses, and the other validation types remain usable for blind internal port scanning through status and timing differences. The caller-supplied apiKey is forwarded to the internal destination as an Authorization Bearer header. A dashboard session is required by default, and none is required when requireLogin is disabled. |
| Koel is a free, open-source music streaming solution. Prior to 9.7.1, outbound podcast and radio fetch paths perform a point-in-time App\Helpers\Network::isPublicHost() or isSafeUrl() check without pinning the validated address, and most paths lack redirect-hop validation and do not revalidate every redirect target. PhanAn\Poddle\Poddle::fromUrl(), PodcastService::getStreamableUrl(), PodcastService::isPodcastObsolete(), App\Rules\HasAudioContentType, and App\Rules\SafeUrl can therefore follow an attacker-controlled redirect to an internal address or connect after DNS rebinding changes a public resolution to a private one. These paths are reachable through podcast and radio APIs, including createPodcastChannel, createInternetRadioStation, refreshPodcasts, apiResource podcasts, and radio/stations, allowing an authenticated user to request internal services or cloud metadata and potentially receive parsed or streamed response content. This issue is fixed in version 9.7.1. |
| A flaw was found in Red Hat Quay. A user with FEATURE_BUILD_SUPPORT enabled and repository write access can exploit a Server-Side Request Forgery (SSRF) vulnerability within the build API. This allows the user to provide a malicious URL, causing the Quay builder to make requests to internal network addresses. Such an action could lead to the disclosure of sensitive internal information. |
| MLflow is an open source AI engineering platform for agents, large language models, and machine learning models. Prior to 3.15.0, the unauthenticated POST /api/2.0/mlflow/webhooks/{id}/test endpoint calls _validate_webhook_url() in mlflow/utils/validation.py only for the original URL while mlflow/webhooks/delivery.py follows redirects and re-resolves the hostname without pinning the validated address, allowing attackers to reach internal or cloud metadata services and receive response_status and response_body. This issue is fixed in version 3.15.0. |
| Plate is a rich-text editor with AI and shadcn/ui. Prior to 53.3.2, @platejs/docx-io fetches remote image URLs while converting attacker-controlled HTML through htmlToDocxBlob in a server-side or privileged environment. The converter can make requests to internal network resources and include the fetched image bytes in the generated DOCX, allowing server-side request forgery with response disclosure. Applications can also incur resource consumption from attacker-selected remote responses. This issue is fixed in version 53.3.2. |
| The unstructured library provides open-source components for ingesting and pre-processing images and text documents, such as PDFs, HTML, Word docs, and many more. From 0.4.7 until 0.24.0, the url argument of partition, partition_html, and partition_md is fetched without host validation in unstructured/partition/auto.py, unstructured/partition/html/partition.py, and unstructured/partition/md.py. An attacker who controls that URL can make a server-side ingestion service request loopback addresses, internal HTTP services, or cloud metadata endpoints through direct targets, redirects, or DNS rebinding. The response body is returned as Element text, allowing internal response disclosure, and side-effecting GET endpoints may also be triggered. This issue is fixed in version 0.24.0. |
| Link Preview JS extracts web links information. Prior to 4.0.4, the resolveDNSHost mitigation in index.ts validates one resolved IP address but fetches the original hostname, allowing an attacker-controlled DNS server to return a public address during validation and a loopback or internal address during the final connection. This DNS rebinding condition bypasses the SSRF protection and can cause the server-side preview fetch to reach internal HTTP resources. Redirect handling is affected by the same validation-to-fetch mismatch. This issue is fixed in version 4.0.4. |
| A Server-Side Request Forgery and supply chain flaw was found in the OpenShift Console Helm catalog proxy. A namespace tenant can plant a ProjectHelmChartRepository with an arbitrary URL that the console pod fetches server-side, bypassing tenant egress restrictions. Combined with catalog metadata poisoning and admin-mediated chart installation, this enables privilege escalation. |
| ATutor is vulnerable to Server-Side request forgery in import functionalities. An authenticated administrator can make the server request arbitrary internal HTTP endpoints, cloud metadata services, or local files via file:// if the PHP
environment permits URL wrappers.
Product is no longer actively supported and the vulnerabilities have not been fixed. Only version 2.2.4 was tested and confirmed as vulnerable, other versions were not tested but might also be vulnerable. |
| n8n before 2.34.1 and 2.33.x before 2.33.4 contains an SSRF protection bypass in the SearXNG Agent tool. The tool sent requests to the user-supplied API URL using a raw HTTP client that did not route through n8n's centralized SSRF protection. On instances with N8N_SSRF_PROTECTION_ENABLED=true, an authenticated user with permission to create SearXNG credentials and configure a personal agent could set the API URL to an internal host, causing the n8n server to connect to that host and return the response content through the Agent chat output. |
| Stigmem before 0.9.0a11 fails to validate the delivery_address parameter when creating webhook subscriptions, allowing authenticated users to specify internal loopback and private network destinations. Attackers can trigger matching fact-change events to cause the Stigmem server to issue server-side HTTP POST requests to internal services, enabling blind SSRF attacks against localhost and private network endpoints. |
| keeper.sh's calendar module version prior to 2.18.14 contains a server-side request forgery (SSRF) guard bypass vulnerability that allows authenticated attackers to reach private network addresses by exploiting a DNS rebinding attack against the two-phase URL validation and connection flow. The SSRF guard validates a hostname's resolved IP addresses but discards them before the actual HTTP connection is opened, allowing an attacker who controls authoritative DNS to return a public address during validation and a private address during the subsequent independent socket-level DNS resolution, causing the guard to pass while the outbound connection reaches internal infrastructure such as cloud instance metadata endpoints. |
