Loading vulnerability…
CVE-2025-24354 affects imgproxy, a server used for resizing, processing, and converting images. When configured with IMGPROXY_ALLOW_LOOPBACK_SOURCE_ADDRESSES set to false, imgproxy is intended to block access to loopback/local addresses as image sources. However, it does not block the IPv4 address 0.0.0.0. An attacker can supply 0.0.0.0 as a source address and cause imgproxy to access services bound to the local host, resulting in a server-side request forgery condition that bypasses the intended loopback protection. The issue is fixed in imgproxy version 3.27.2.
Mallory correlates every CVE against your assets, your vendors, and active adversary campaigns. Know which vulnerabilities matter for you, not just which ones are loud.
What it means. What to do now. Patch path, mitigations, and the assume-compromise checklist.
What an attacker gets, and what they’ve been doing with it.
If you can’t patch tonight, do this now.
Patch, then assume compromise.
1 valid exploit after Mallory filtered fakes, detection scripts, and README-only repos.
This repository contains a single Python script (CVE-2025-24354.py) that exploits an SSRF (Server-Side Request Forgery) vulnerability in imgproxy (CVE-2025-24354). The script is a proof-of-concept tool that automates the enumeration of internal network services accessible from the imgproxy server. It does so by crafting requests to the imgproxy service, causing it to fetch URLs pointing to various internal ports and schemes (http, https) on 0.0.0.0 (localhost). The script supports customization of ports, schemes, and paths, and can output results to the console or a CSV file. It reports HTTP response codes, content types, and truncated response bodies for each attempted endpoint, helping an attacker map out internal services reachable via the SSRF flaw. The script is standalone, requires the 'requests' library, and is invoked from the command line with the target imgproxy server specified. No hardcoded credentials or external endpoints are present; all targeting is dynamic based on user input and internal network enumeration.
5 sources tracked across advisories and community write-ups. News coverage will land here when it surfaces.
No news coverage yet. Advisories and community discussion only.
Query your assets running an affected version, and investigate the blast radius.
Every observed campaign linking this CVE to a named adversary.
Malware families riding this exploit, with evidence and IOCs.
YARA, Sigma, Snort, and vendor rules, auto-deployed to your SIEM.
Cross-references every affected SKU, including bundled OEM variants.
Community discussion across Reddit, Mastodon, and other social sources.