| タイトル | houtini-ai seo-crawler-mcp 2.1.5 Server‑Side Request Forgery |
|---|
| 説明 | A Server-Side Request Forgery (SSRF) vulnerability exists in the run_seo_audit tool of houtini-ai/seo-crawler-mcp, where user-controlled URLs are passed directly to the Crawlee HTTP crawler without validation (src/core/CrawlOrchestrator.ts:147). The vulnerable data flow begins when an MCP client sends a request with a malicious URL parameter, which passes through src/index.ts:173 and src/tools/run-seo-audit.ts:30 before triggering an HTTP request at the crawler execution point. Dynamic testing with MCP-Inspector confirmed exploitability by successfully accessing a localhost web service on port 8001, with response content saved to the local SQLite database. This vulnerability (CVSS 7.5 HIGH) enables attackers to access internal services, scan private networks (10.x.x.x, 172.16.x.x, 192.168.x.x), steal cloud metadata credentials from x.x.x.x, and bypass firewall restrictions. In cloud deployments, attackers can retrieve IAM role credentials leading to complete account compromise. CodeQL static analysis failed to detect this vulnerability because HTTP requests are made through the third-party Crawlee library rather than direct fetch() calls, with complex multi-layer data flow making taint tracking difficult. Remediation requires implementing URL validation to block localhost addresses, private IP ranges, and cloud metadata endpoints before passing user input to the crawler, affecting all versions up to commit 18ca435a3437d3c3699a10a42976d96667fef732. |
|---|
| ソース | ⚠️ https://github.com/Net-defense/public_exp/issues/1 |
|---|
| ユーザー | Lululu (UID 101009) |
|---|
| 送信 | 2026年09月04日 05:52 (2 月 ago) |
|---|
| モデレーション | 2026年10月11日 10:34 (1 month later) |
|---|
| ステータス | 承諾済み |
|---|
| VulDBエントリ | 416565 [houtini-ai seo-crawler-mcp 迄 2.1.5 Crawler CrawlOrchestrator.ts run_seo_audit 特権昇格] |
|---|
| ポイント | 20 |
|---|