CVE Watch19 oct 2026

CVE-2026-112267

SGLang through 0.5.22 in prefill/decode disaggregation mode contains a server-side request forgery vulnerability that allows unauthenticated

CVSS

6.5

Medio

EPSS

—

KEV

—

Exploit Today

0-100

Publicado: 19 oct 2026 · Última mod.: 19 oct 2026 · CWE-918

EPSS · 30d

Sin historial EPSS suficiente todavía.

Descripción técnica

SGLang through 0.5.22 in prefill/decode disaggregation mode contains a server-side request forgery vulnerability that allows unauthenticated attackers to trigger outbound requests via client-supplied bootstrap_host and bootstrap_port fields. Attackers sending crafted requests to the Decode worker's /generate endpoint can probe internal hosts and ports and block the shared scheduler loop, delaying other inference requests.

Referencias oficiales

CVEs relacionados

CVECVSSEPSSKEVRExplotTítuloVis.

CVE-2026-1117615.8 MED

—

———A2A Java SDK is the official Java SDK for the Agent2Agent (A2A) Protocol. Prior to 1.3.0.Final, BasePushNotificationSender in server-common sends task data to the URL in TaskPushNotificationConfig without destination validation, scheme restrictions, or bounds on redirects. InMemoryPushNotificationConfigStore.setInfo stores the URL verbatim, and onCreateTaskPushNotificationConfig does not validate the destination. In the default deployment with push notifications enabled, an unauthenticated caller can register a destination through an inline SendMessage configuration or CreateTaskPushNotificationConfig for an existing task, causing task status and artifact updates to send requests from the server's network position to loopback, private, link-local cloud-metadata, IPv6 unique-local, unspecified, or multicast destinations where routing permits. JdkA2AHttpClient and VertxA2AHttpClient follow redirects, allowing a webhook redirect to bypass creation-time destination checks and convert a POST into a GET; hostname-only checks also do not prevent DNS rebinding. Direct POST delivery can push attacker-influenced task data into internal services, whereas a redirected POST-to-GET request drops the body. The sender discards the fetched response and does not expose delivery success or failure, so the demonstrated behavior does not provide response read-back. No credentials or user interaction are required. This issue is fixed in version 1.3.0.Final.7h CVE-2026-1059418.1 ALT

—

———The Azure Key Vault KMS provider in Kroxylicious accepts an untrusted vaultName from wire-deserialized EDEK bytes without format validation, enabling SSRF that exfiltrates the proxy's Azure bearer token to an attacker-controlled host. Token replay against the real Key Vault defeats the entire record-encryption trust model.7h CVE-2026-107597—

—

———PIA's token verification (`verify_token` in `pia/oidc.py`) fetches the OIDC discovery document of an allowlisted issuer and then follows the `jwks_uri` contained in that response without parsing it as a URL, comparing it against the approved issuer's origin, or requiring HTTPS — the only check is that the field is non-empty (`pia/oidc.py:42-52`). An approved identity provider — and every registered per-project Jenkins instance under `https://ci.eclipse.org/<project>/oidc` is a separate such party (`pia/models.py`) — can therefore direct PIA to issue a GET request to an arbitrary host and port, including over plaintext `http`, from PIA's own network position. The fetch happens before any signature is verified, as a side effect of an unauthenticated `POST /v1/upload/sbom` request that is ultimately rejected with 401.11h CVE-2026-112237—

—

———The extension provides a public proxy endpoint that fetches a URL on behalf of the visitor. For one HTTP request method, the endpoint does not apply the authorization check that protects the other methods, so the URL it fetches is fully controlled by the requester. An unauthenticated attacker can make the server send requests to arbitrary internal or external hosts. The attacker receives the status code and selected response headers, which allows probing of internal hosts and ports. Response bodies are not disclosed.12h CVE-2026-83555.4 MED

—

——0IBM ContextForge MCP Gateway 0.1.0 through 1.0.5 is vulnerable to a server-side request Forgery (SSRF) that allows authenticated users with gateway creation privileges to perform arbitrary HTTP POST requests to internal network resources and cloud metadata endpoints. The vulnerability occurs during OAuth-enabled gateway registration when the system fails to validate the oauth_config.token_url parameter. When registering a gateway with OAuth client credentials authentication, an attacker can supply a malicious token_url pointing to internal services, cloud metadata endpoints (AWS IMDSv1, GCP, Azure), or localhost-bound services. The application immediately performs an HTTP POST request to the attacker-controlled URL with form-encoded credentials in the body. If the response cannot be parsed as a valid OAuth token response, up to 256 characters of the response body are reflected back to the attacker in the HTTP 502 error message. This vulnerability bypasses existing SSRF protections that are applied to the gateway.url field but not to OAuth configuration parameters.7h CVE-2026-112991—

—

——0The html_to_markdown expansion module in MISP Modules contains a server-side request forgery (SSRF) vulnerability in its is_safe_url() URL validation function. The BLOCKED_RANGES list used to reject unsafe destination addresses was incomplete, omitting several critical IP address ranges. Specifically, the following ranges were not blocked: 0.0.0.0/8 (unspecified, which routes to loopback on Linux), 100.64.0.0/10 (carrier-grade NAT per RFC 6598), ::/128 (IPv6 unspecified), fc00::/7 (IPv6 Unique Local Addresses), and fe80::/10 (IPv6 link-local). The existing list only covered IPv4 loopback (127.0.0.0/8), RFC 1918 private ranges, IPv4 link-local (169.254.0.0/16), and IPv6 loopback (::1/128). An authenticated user with the ability to create or modify MISP attributes that trigger the html_to_markdown expansion module could supply URLs pointing to these unblocked internal address ranges. Because the module fetches the URL server-side to convert HTML content to Markdown, the request is made from the MISP server's network context. This allows an attacker to probe and read responses from internal services, carrier-grade NAT infrastructure, or link-local endpoints that would otherwise be unreachable from external networks. IPv4-mapped IPv6 forms (e.g., [::ffff:0.0.0.0]) also bypassed the existing checks. The security impact is primarily confidentiality: an attacker can read data served by internal services reachable from the MISP server. The MISP instance itself is not compromised, but it is abused as a proxy to access otherwise restricted network resources.15h