Loading vulnerability…
Squidbleed is an out-of-bounds read vulnerability in Squid’s FTP gateway affecting versions prior to 7.6. The flaw is in FTP directory-listing parsing in src/clients/FtpGateway.cc, where syntactic validation is insufficient when a TypeA or TypeB listing entry contains a date or timestamp that is not followed by a filename. In that condition, parsing is not properly constrained to the input buffer, allowing the parser to advance past the terminating boundary and read adjacent heap memory. The disclosed root cause is a missing end-of-string guard in the whitespace-skipping logic before searching for the next non-whitespace character. As a result, memory from unrelated transactions can be copied into the generated gateway output and returned to the requesting client. The issue is particularly relevant in shared proxy deployments because reused buffers may contain fragments of other users’ HTTP requests and associated sensitive application data.
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 is a compact standalone proof-of-concept exploit for CVE-2026-47729 ('Squidbleed'), targeting Squid's FTP handling to trigger an information disclosure condition. The repository contains only two files: a README and a single Python exploit script, making CVE-2026-47729.py the clear entry point. The exploit has two tightly integrated components in one script. First, it starts an attacker-controlled FTP server that emulates enough FTP behavior to satisfy a client: USER/PASS, SYST, PWD, TYPE, EPSV, LIST/NLST, and QUIT. The key malicious behavior is in the LIST/NLST handling, where it sends a crafted truncated directory listing line and closes the data connection, intended to trigger the vulnerable memory over-read behavior in Squid. Second, the script acts as a poller/harvester against a target Squid proxy. It repeatedly opens a TCP connection to the configured proxy (default 127.0.0.1:3128) and sends an HTTP GET request for an ftp:// URL pointing to the attacker FTP server (default ftp://anon:x@127.0.0.1:2222/). It then reads the proxy response body and searches for leaked data embedded in HTML href content. The script URL-decodes the leaked bytes and applies regex extraction for Basic and Bearer tokens. Basic tokens are additionally Base64-decoded and printed as username:password when possible. Operationally, the exploit is multithreaded: one background thread runs the FTP server, multiple worker threads continuously poll the proxy, and a status thread reports polling rate, hit count, and distinct token counts. This is not merely a detector; it actively attempts exploitation and harvests sensitive material from leaked memory. There is no post-exploitation shell or code execution payload—its purpose is credential and token disclosure from a vulnerable Squid instance.
Products and vendors Mallory has correlated with this vulnerability. Open in Mallory to drill down to specific CPE configurations and version ranges.
Vendor-confirmed product mapping. Mallory continuously reconciles this list against your asset inventory.
105 sources tracked across advisories, community write-ups, and news. New activity surfaces here as Mallory finds it.
A Squid vulnerability cited as an example of shared attribution across multiple AI-assisted discovery paths; no technical details are provided.
A moderate-severity information disclosure vulnerability in Squid where a malicious FTP response can interfere with HTTP gateway output and leak memory from the proxy worker.
Out-of-bounds read / information disclosure vulnerability in Squid's FTP gateway parser that can leak adjacent heap memory, including Authorization headers, cookies, session tokens, and API keys.
A heap over-read vulnerability in Squid that can leak cleartext HTTP data; referred to as 'Squidbleed' and notable for affecting long-standing code.
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.