Dozzle CVE-2026-102332

Exploitation requires ALL of: (1) attacker ability to control container display names/labels reported to Dozzle - either directly on the Docker engine or via a Dozzle agent that forwards container names; (2) a legitimate Dozzle user actively invoking the log-download endpoint and extracting the resulting ZIP archive (UI:A); and (3) an extraction tool that does not itself guard against path-traversal entry names. The vulnerable versions are Dozzle before 11.1.2. Because the write occurs on the extracting user's machine (subsequent-system integrity, SI:H) rather than on the Dozzle server, and depends on all three conditions aligning, realistic exploitability is significantly narrower than the network attack vector alone suggests.

This is a Zip Slip (path-traversal-via-archive) flaw in Dozzle's log-download endpoint, not a directly server-compromising bug. The vendor CVSS 4.0 vector (AV:N/AC:L/AT:P/PR:N/UI:A/SI:H) scores 4.6 (Medium) and is internally consistent: the vulnerable Dozzle system itself is untouched (VC:N/VI:N/VA:N) - the integrity impact lands on a SUBSEQUENT system (SI:H), namely the filesystem of the user who downloads and extracts the malicious ZIP. AT:P (attack requirements present) and UI:A (active user interaction) correctly capture that exploitation is not turn-key: the attacker must be able to label/name containers AND a legitimate user must download and extract the log archive with a path-traversal-unsafe extractor. No EPSS score, no CISA KEV listing, and no public exploit code (the patch ships only a unit test asserting entry-name sanitization). There is no active-exploitation signal at all. One point worth verifying: the vendor rated PR:N, but the description's precondition 'attackers who can label containers' implies some level of control over the Docker/agent environment, which arguably warrants PR:L - this is deployment-dependent (e.g., a compromised Dozzle agent or a tenant able to push labeled containers). Overall a real-but-narrow, low-to-medium priority issue: patch on a normal cadence rather than as an emergency.

An attacker with the ability to name or label containers in a monitored Docker environment (e.g., via a compromised Dozzle agent, or by deploying a container in a shared/multi-tenant host) creates a container named something like '../../etc/cron.d/evil'. When an operator later uses Dozzle's log-download feature and extracts the resulting ZIP with a naive extraction tool, the crafted entry name escapes the extraction directory and writes an attacker-controlled file to an arbitrary path, enabling code execution. No standalone public exploit exists, though the vendor's fix includes a unit test demonstrating the '../../etc/cron.d/evil' payload.