CVE-2026-92705: 
Linux Debian vulnerability analysis and mitigation

CVE-2026-92705 is an arbitrary code execution vulnerability in Aegisub, a cross-platform advanced subtitle editor, affecting versions 3.2.0 through 3.4.2. The flaw stems from two related weaknesses: Aegisub automatically loads Automation scripts referenced in ASS subtitle project metadata without user consent (CWE-829), and versions 3.4.0–3.4.2 additionally suffer from inconsistent NUL character handling that enables a single-file ASS/Lua polyglot attack (CWE-158). It was published on October 9, 2026, and carries a CVSS v3.1 base score of 7.8 (High) (GitHub Advisory, Red Hat Bugzilla).

The primary root cause (CWE-829) is that Aegisub unconditionally loads Lua Automation scripts listed in the Automation Scripts metadata section of ASS subtitle files upon opening, without prompting the user for authorization. A secondary issue (CWE-158) present in versions 3.4.0–3.4.2 involves the agi::fs::HasExtension function evaluating extension validity against the full string including characters after an embedded NUL byte, while filesystem operations (which use C-string semantics) truncate at the NUL. This discrepancy allows a crafted ASS file to embed a NUL in its filename reference so that extension validation sees .lua while the filesystem resolves the path to the ASS file itself — enabling a single-file ASS/Lua polyglot that loads and executes itself as a Lua script. The fix in commit 0a3073d normalizes HasExtension to use C-string semantics, commit b5bf239 sanitizes NUL bytes in ASS project metadata parsing, and commit e409905 adds an explicit user trust prompt before loading any referenced local Automation scripts (GitHub Advisory, Commit e409905, Commit 0a3073d, Commit b5bf239).

Successful exploitation results in arbitrary code execution with the full privileges of the Aegisub process on the victim's system, compromising confidentiality, integrity, and availability of user data. An attacker can read, modify, or delete files accessible to the user, install persistent malware, or use the compromised host as a pivot point for further lateral movement within a network. No authentication or elevated privileges are required on the attacker's side; the only prerequisite is that the victim opens a malicious ASS subtitle file (GitHub Advisory).

No public proof-of-concept exploit code or in-the-wild exploitation has been observed as of the time of publication (Feedly). The EPSS score is approximately 0.0013 (0.13%), reflecting a low near-term exploitation probability. The vulnerability is not listed in the CISA Known Exploited Vulnerabilities (KEV) catalog. Exploitation requires user interaction — specifically, a victim must open a crafted ASS file — but no privileges are required from the attacker, making social engineering (e.g., distributing subtitle files via fan-subbing communities or torrent sites) a plausible delivery vector (GitHub Advisory, Red Hat Bugzilla).

1. Craft a malicious Automation script: Create a Lua script (malicious.lua) containing arbitrary code to execute on the victim's machine (e.g., a reverse shell, file exfiltration command, or dropper).
2. Craft a malicious ASS subtitle file: Create or modify an ASS subtitle file to include an Automation Scripts metadata line referencing the malicious Lua script, e.g., Automation Scripts: malicious.lua in the [Aegisub Project Garbage] section.
3. Bundle and distribute: Package the crafted ASS file alongside the malicious.lua script (e.g., in a ZIP archive or torrent) and distribute it through subtitle-sharing sites, fan-subbing communities, or direct messaging.
4. Victim opens the file: When the victim opens the ASS file in Aegisub 3.2.0–3.4.2, the application automatically reads the Automation Scripts metadata and loads malicious.lua without any trust prompt.
5. Code executes: The Lua script executes with the privileges of the Aegisub process, achieving the attacker's objective (e.g., establishing persistence, exfiltrating data, or deploying additional payloads).
6. (Variant for 3.4.0–3.4.2 — single-file polyglot): Craft a single file that is simultaneously a valid ASS subtitle file and a valid Lua script. Embed a NUL byte in the Automation Scripts metadata path so that extension validation sees .lua but the filesystem resolves the path to the ASS file itself. Distributing only this single file is sufficient for exploitation (GitHub Advisory).

• File System: Unexpected .lua files co-located with ASS subtitle files in download or working directories; ASS files containing Automation Scripts: metadata lines referencing local paths; ASS files with embedded NUL bytes (visible in binary-safe hex editors) in metadata sections.
• Process: Unusual child processes spawned by the Aegisub process (e.g., cmd.exe, powershell.exe, bash, curl, wget, python) shortly after opening a subtitle file; network connections initiated by the Aegisub process to external IPs.
• Logs: Application or system logs showing Aegisub loading Lua scripts from non-standard or user-writable directories; OS-level audit logs recording file reads of .lua files triggered by the Aegisub process.
• Network: Outbound connections from the Aegisub process to unexpected remote hosts, particularly on common reverse-shell ports (e.g., 4444, 1337, 8080) (GitHub Advisory).

Users should upgrade to Aegisub 3.5.0, which addresses the vulnerability through three commits: adding a user trust prompt before loading local Automation scripts (e409905), making HasExtension consistent with filesystem NUL-termination semantics (0a3073d), and sanitizing NUL bytes in ASS project metadata parsing (b5bf239). Until upgrading, users should avoid opening ASS subtitle files or subtitle project bundles from untrusted sources. As a partial manual workaround, inspect ASS files with a binary-safe text editor before opening and remove all Automation Scripts: metadata lines, ensuring the editor exposes embedded NUL characters. Note that disabling automatic loading of linked audio/video files does not mitigate this vulnerability, as Automation scripts are loaded through a separate mechanism (GitHub Advisory, Red Hat Bugzilla).

The vulnerability was reported by shavitush, with remediation developed by arch1t3cht and coordinated by CoffeeFlux under the GitHub Security Advisory program. Coverage appeared on Linux security news outlets including Pro-Linux.de and LinuxSecurity.com, and the CVE was noted on Mastodon security feeds shortly after publication. No major vendor statements beyond the upstream GitHub advisory and Red Hat Bugzilla tracking have been observed (GitHub Advisory, Red Hat Bugzilla).