Wall of Fame

Vulnerabilities discovered by Tachyon that are making the internet safer. Every CVE represents real threats we helped neutralize.

high
CVSS 8.2

CVE-2026-28416

SSRF via Malicious proxy_url Injection in gr.load() Config Processing

A Server-Side Request Forgery vulnerability in Gradio allows attackers to conduct arbitrary HTTP requests through victim servers. When applications use gr.load() to import external Spaces, untrusted proxy_url fields from remote Spaces are automatically added to an allowlist without validation, enabling access to cloud metadata endpoints, internal databases, and private network infrastructure.

Gradio <= 6.5.1
Published 2026-02-27
medium
CVSS 4.3

CVE-2026-28415

Open Redirect in OAuth Flow

A validation flaw in Gradio's OAuth implementation allows attackers to redirect users to arbitrary external sites through an unvalidated _target_url parameter in the /logout and /login/callback endpoints. Attackers can craft malicious URLs that exploit users' trust in the legitimate hf.space domain to conduct phishing attacks.

Gradio <= 6.5.1
Published 2026-02-27
critical
CVSS 9.3

GHSA-qrq5-wjgg-rvqw

Path Traversal in Plugin Installation

A path traversal issue in OpenClaw plugin installation allows a malicious plugin package name to escape the intended extensions directory and write files outside of it, enabling arbitrary file writes on the system.

OpenClaw >= 2026.1.29-beta.1
Published 2026-02-14
medium
CVSS 5.3

CVE-2026-25738

Server-Side Request Forgery (SSRF) in Multiple Places

Indico makes outgoing requests to user-provided URLs in various places without properly blocking access to localhost or cloud metadata endpoints, allowing SSRF attacks to exfiltrate sensitive data.

Indico < 3.3.10
Published 2026-02-17
critical
CVSS 9.1

CVE-2026-25227

Authenticated RCE via Policy/Property Mapping Test Endpoint

Users with "Can view Property Mapping" or "Can view Expression Policy" permissions can exploit the test endpoint to run arbitrary code within the authentik container, accessing the entire database and environment variables.

authentik >= 2021.3.1
Published 2026-02-12
critical
CVSS 9.4

CVE-2026-24780

Remote Code Execution via Disabled Block Execution

A critical authorization flaw in AutoGPT Platform allows authenticated users to bypass security controls and achieve remote code execution by exploiting endpoints that fail to validate the disabled flag on blocks.

AutoGPT Platform v0.1.0+
Published 2026-01-29
high
CVSS 8.1

CVE-2025-14297

Authorization Bypass in MLflow Basic Auth

Authenticated non-admin users can bypass per-object authorization and access or modify restricted resources via unprotected Flask routes and GraphQL endpoints when basic-auth is enabled.

MLflow v2.3.2
Published 2025-08-21
high
CVSS 7.4

CVE-2026-24123

Path Traversal via Bentofile Configuration

Attackers can craft a malicious bentofile that exfiltrates arbitrary files from the filesystem into the bento archive through insufficient path validation in configuration file processing.

BentoML ≤1.4.33
Published 2026-01-26