GHSA-W6X2-F5M4-XW82

Vulnerability from github – Published: 2026-11-19 22:41 – Updated: 2026-11-19 22:41
VLAI
Summary
fzf: Arbitrary command execution via bash arithmetic expansion on unvalidated layout size in fzf-tmux (-l/-u attached value)
Details

Summary

bin/fzf-tmux fails to validate layout size values given in the attached form (e.g. -l50). For -l/-u these values reach bash arithmetic expansion ($(( max - size ))), where an operand like a[$(cmd)] executes cmd as the invoking user during option parsing — before any tmux interaction and even outside tmux. A second sink embeds unvalidated popup geometry (-x/-y/-w/-h/-p) into an unquoted tmux popup ... $opt ... invocation, allowing injection of tmux flags such as -e BASH_ENV=<path> (command execution via non-interactive bash sourcing $BASH_ENV).

Reporter / Credits:

Sai Sreewathsa Kovalluri, Shon Babu and Kannan K (Researchers), Innspark Solutions https://innspark.in/

Details

In bin/fzf-tmux (tag 0.74.4, commit a140afeb), the option parser validates size values with a regex (^[0-9%,]+$ / ^[A-Z]$) only for the separate-argument form (line 82). The attached form is captured unvalidated at line 80:

if [[ ${#arg} -gt 2 ]]; then
  size="${arg:2}"                       # no validation
else
  if [[ $1 =~ ^[0-9%,]+$ ]] || [[ $1 =~ ^[A-Z]$ ]]; then   # only separate form validated

For -l and -u, swap is non-empty (lines 68/75), so size reaches arithmetic evaluation:

  • line 120: size=$((max - size)) (non-percent values)
  • line 102: opt="$opt -l $((100 - size))%" (percent, tmux >= 3.2)
  • line 108: opt="$opt -p $((100 - size))" (percent, tmux < 3.2)

Bash arithmetic expands array subscripts, so a[$(command)] runs command. Verified end-to-end against the unmodified script (see PoC).

Second sink (same root cause): opt is used unquoted at line 236 (tmux popup -d "$PWD" $opt "bash $argsf") and line 249, so a geometry value containing spaces injects extra tmux flags, e.g. -e BASH_ENV=/path/to/script.

PoC

Verified on 0.74.4 (no tmux or fzf binary needed — a stub fzf on PATH suffices because the sink fires during bash argument parsing):

$ bash bin/fzf-tmux '-l a[$(id > /tmp/pwn1)]'      # line 120 sink (non-percent)
$ bash bin/fzf-tmux '-u a[$(id > /tmp/pwn2)]%'     # line 102/108 sink (percent)

$ cat /tmp/pwn1 /tmp/pwn2
uid=1001(sb) gid=1001(sb) groups=...
uid=1001(sb) gid=1001(sb) groups=...

A self-contained reproducer was provided by email (poc-t03-arithmetic.sh). Popup flag-injection variant (tmux >= 3.2): bash bin/fzf-tmux -p '-x0 -e BASH_ENV=/path/to/evil.sh'.

Impact

Local command injection (CWE-78). fzf-tmux argv is commonly assembled by third-party scripts and editor plugins (e.g. fzf-tmux -l "${height}%"); any layout value influenced by less-trusted data (project-local config, wrapper scripts) yields arbitrary command execution as the invoking user. Suggested fixes: apply the regex validation to the attached form, constrain size to [0-9]+ before any $((...)), and build the tmux invocation as an array ("${opt_parts[@]}") instead of unquoted $opt.

Reported privately to the maintainer by email; fixes prepared for the next release.

Exploitability in practice

Impact is high (command execution as the invoking user), but a real attack requires attacker-controlled data to reach the fzf-tmux layout arguments (-l/-u). In normal use these values are set by the user or by trusted, hardcoded configuration, so a vulnerable data flow is uncommon. The issue is local-only, requires the victim to run fzf-tmux with the poisoned value, and bin/fzf-tmux has been superseded by the built-in --tmux/--popup option since 0.53.0.

Show details on source website

{
  "affected": [
    {
      "database_specific": {
        "last_known_affected_version_range": "\u003c= 0.74.4"
      },
      "package": {
        "ecosystem": "Go",
        "name": "github.com/junegunn/fzf"
      },
      "ranges": [
        {
          "events": [
            {
              "introduced": "0"
            },
            {
              "fixed": "0.74.5"
            }
          ],
          "type": "ECOSYSTEM"
        }
      ]
    }
  ],
  "aliases": [
    "CVE-2026-121716"
  ],
  "database_specific": {
    "cwe_ids": [
      "CWE-78"
    ],
    "github_reviewed": true,
    "github_reviewed_at": "2026-11-19T22:41:46Z",
    "nvd_published_at": "2026-11-17T20:17:36Z",
    "severity": "HIGH"
  },
  "details": "### Summary\n\n`bin/fzf-tmux` fails to validate layout size values given in the *attached* form (e.g. `-l50`). For `-l`/`-u` these values reach bash arithmetic expansion (`$(( max - size ))`), where an operand like `a[$(cmd)]` executes `cmd` as the invoking user during option parsing \u2014 before any tmux interaction and even outside tmux. A second sink embeds unvalidated popup geometry (`-x/-y/-w/-h/-p`) into an unquoted `tmux popup ... $opt ...` invocation, allowing injection of tmux flags such as `-e BASH_ENV=\u003cpath\u003e` (command execution via non-interactive bash sourcing `$BASH_ENV`).\n\n### Reporter / Credits:\nSai Sreewathsa Kovalluri, Shon Babu and Kannan K (Researchers), Innspark Solutions https://innspark.in/\n\n### Details\n\nIn `bin/fzf-tmux` (tag 0.74.4, commit a140afeb), the option parser validates size values with a regex (`^[0-9%,]+$` / `^[A-Z]$`) only for the separate-argument form (line 82). The attached form is captured unvalidated at line 80:\n\n```bash\nif [[ ${#arg} -gt 2 ]]; then\n  size=\"${arg:2}\"                       # no validation\nelse\n  if [[ $1 =~ ^[0-9%,]+$ ]] || [[ $1 =~ ^[A-Z]$ ]]; then   # only separate form validated\n```\n\nFor `-l` and `-u`, `swap` is non-empty (lines 68/75), so `size` reaches arithmetic evaluation:\n\n- line 120: `size=$((max - size))` (non-percent values)\n- line 102: `opt=\"$opt -l $((100 - size))%\"` (percent, tmux \u003e= 3.2)\n- line 108: `opt=\"$opt -p $((100 - size))\"` (percent, tmux \u003c 3.2)\n\nBash arithmetic expands array subscripts, so `a[$(command)]` runs `command`. Verified end-to-end against the unmodified script (see PoC).\n\nSecond sink (same root cause): `opt` is used unquoted at line 236 (`tmux popup -d \"$PWD\" $opt \"bash $argsf\"`) and line 249, so a geometry value containing spaces injects extra tmux flags, e.g. `-e BASH_ENV=/path/to/script`.\n\n### PoC\n\nVerified on 0.74.4 (no tmux or fzf binary needed \u2014 a stub `fzf` on PATH suffices because the sink fires during bash argument parsing):\n\n```console\n$ bash bin/fzf-tmux \u0027-l a[$(id \u003e /tmp/pwn1)]\u0027      # line 120 sink (non-percent)\n$ bash bin/fzf-tmux \u0027-u a[$(id \u003e /tmp/pwn2)]%\u0027     # line 102/108 sink (percent)\n\n$ cat /tmp/pwn1 /tmp/pwn2\nuid=1001(sb) gid=1001(sb) groups=...\nuid=1001(sb) gid=1001(sb) groups=...\n```\n\nA self-contained reproducer was provided by email (poc-t03-arithmetic.sh). Popup flag-injection variant (tmux \u003e= 3.2): `bash bin/fzf-tmux -p \u0027-x0 -e BASH_ENV=/path/to/evil.sh\u0027`.\n\n### Impact\n\nLocal command injection (CWE-78). `fzf-tmux` argv is commonly assembled by third-party scripts and editor plugins (e.g. `fzf-tmux -l \"${height}%\"`); any layout value influenced by less-trusted data (project-local config, wrapper scripts) yields arbitrary command execution as the invoking user. Suggested fixes: apply the regex validation to the attached form, constrain `size` to `[0-9]+` before any `$((...))`, and build the tmux invocation as an array (`\"${opt_parts[@]}\"`) instead of unquoted `$opt`.\n\nReported privately to the maintainer by email; fixes prepared for the next release.\n\n### Exploitability in practice\n\nImpact is high (command execution as the invoking user), but a real attack requires attacker-controlled data to reach the fzf-tmux layout arguments (`-l`/`-u`). In normal use these values are set by the user or by trusted, hardcoded configuration, so a vulnerable data flow is uncommon. The issue is local-only, requires the victim to run fzf-tmux with the poisoned value, and `bin/fzf-tmux` has been superseded by the built-in `--tmux`/`--popup` option since 0.53.0.",
  "id": "GHSA-w6x2-f5m4-xw82",
  "modified": "2026-11-19T22:41:46Z",
  "published": "2026-11-19T22:41:46Z",
  "references": [
    {
      "type": "WEB",
      "url": "https://github.com/junegunn/fzf/security/advisories/GHSA-w6x2-f5m4-xw82"
    },
    {
      "type": "ADVISORY",
      "url": "https://nvd.nist.gov/vuln/detail/CVE-2026-121716"
    },
    {
      "type": "WEB",
      "url": "https://github.com/junegunn/fzf/commit/f126ff29549e0983f571fbc32dddcfbede251aa7"
    },
    {
      "type": "PACKAGE",
      "url": "https://github.com/junegunn/fzf"
    },
    {
      "type": "WEB",
      "url": "https://github.com/junegunn/fzf/releases/tag/v0.74.5"
    }
  ],
  "schema_version": "1.4.0",
  "severity": [
    {
      "score": "CVSS:3.1/AV:L/AC:H/PR:N/UI:R/S:U/C:H/I:H/A:H",
      "type": "CVSS_V3"
    }
  ],
  "summary": "fzf: Arbitrary command execution via bash arithmetic expansion on unvalidated layout size in fzf-tmux (-l/-u attached value)"
}



Log in or create an account to share your comment.




Tags
Taxonomy of the tags.


Loading…

Loading…

Loading…

Forecast uses a logistic model when the trend is rising, or an exponential decay model when the trend is falling. Fitted via linearized least squares.

Sightings

Author Source Type Date Other

Nomenclature

  • Seen: The vulnerability was mentioned, discussed, or observed by the user.
  • Confirmed: The vulnerability has been validated from an analyst's perspective.
  • Published Proof of Concept: A public proof of concept is available for this vulnerability.
  • Exploited: The vulnerability was observed as exploited by the user who reported the sighting.
  • Patched: The vulnerability was observed as successfully patched by the user who reported the sighting.
  • Not exploited: The vulnerability was not observed as exploited by the user who reported the sighting.
  • Not confirmed: The user expressed doubt about the validity of the vulnerability.
  • Not patched: The vulnerability was not observed as successfully patched by the user who reported the sighting.

Loading…

Loading…

Loading…

Related by attack behaviour

Vulnerabilities whose description is nearest to this one in the vector space of the CIRCL/vulnerability-attack-technique-biencoder model. This is a similarity search over the bi-encoder space (plain cosine), not a classification, and it has no measured accuracy.


Loading…