CVE-2026-86404

InternLM LMDeploy 0.9.0 through 0.18.0 contains an unauthenticated remote code execution vulnerability. Attackers can send a base64-encoded crafted pickle in serialized_named_tensors to the /update_weights endpoint, which ForkingPickler.loads deserializes without validation, executing commands as the api_server process.

A deserialization of untrusted data vulnerability in the SubWorkflow task in Google Cloud Application Integration versions prior to 2026-09-11 allows an authenticated attacker to execute arbitrary code using a maliciously crafted serialized object. The vulnerability was patched on 11 September 2026.

In SOCFortress CoPilot prior to 0.1.77, the POST /api/scoutsuite/generate-gcp-report endpoint passes the raw multipart filename to the file writer without sanitization or a resolved-path containment check. A traversal filename therefore writes the supplied JSON outside the intended report directory, including into the application source directory.

In the ng-antd-admin template prior to 22.0, a hardcoded default JWT secret permits token forgery when a deployment has not replaced the default secret, without a startup warning. An attacker can bypass authentication and authorization and impersonate any user, including administrators.

In SOCFortress CoPilot, the unauthenticated /scoutsuite-report StaticFiles mount exposes the scoutsuite-report directory, disclosing HTML reports and result files containing AWS, Azure and GCP IAM principals, resource inventories and misconfigurations. During a GCP scan, the service-account key is saved into that same web-served directory and removed only after the scan finishes, so an anonymous client can retrieve it.

In SOCFortress CoPilot prior to 0.1.76, verify_graylog_header and verify_velociraptor_header authenticate webhooks against the same publicly hardcoded fallback header, also shipped in .env.example. Deployments with unchanged secrets let an anonymous caller bypass JWT authentication, invoke active-response actions and create forged alerts.

In SOCFortress CoPilot, portal users including customer_user accounts can submit stored markup via the alert comment endpoint, which is rendered without sanitization using v-html. This enables stored XSS, and because access and refresh tokens are persisted in JavaScript-readable storage, injected script can steal them and impersonate the viewer. The same unsanitized rendering affects other SOC fields (cases, assets, health-checks), with fixes first contained in versions 0.1.70 (health-check) and 0.1.76 (comments and other SOC sinks).

MuseBot before 2.0.0 contains an authentication bypass vulnerability that allows unauthenticated attackers to forge admin sessions using the hard-coded default session_key 'muse_bot_session_key'. Attackers can sign a session cookie carrying a user_id to gain full administrative access.

In Vvveb prior to version 1.0.8.7, an authenticated IDOR on the admin/auth-token endpoint allows any administrator to read and overwrite another administrator's API tokens. The issue has been patched in version 1.0.8.7.

In Vvveb prior to version 1.0.8.7, the password reset functionality is vulnerable to an authentication bypass due to insufficient validation of the user and token GET parameters. The issue has been patched in version 1.0.8.7.