CVE-2023-46229 is a server-side request forgery vulnerability in LangChain versions before 0.0.317. RecursiveUrlLoader did not sufficiently validate externally supplied crawl target URLs or subsequent URLs encountered during recursive crawling. An attacker-controlled external crawl target could direct the loader to request internal network services or other restricted endpoints.
Mallory correlates every CVE against your assets, your vendors, and active adversary campaigns. Know which vulnerabilities matter for you, not just which ones are loud.
What it means. What to do now. Patch path, mitigations, and the assume-compromise checklist.
What an attacker gets, and what they’ve been doing with it.
If you can’t patch tonight, do this now.
Patch, then assume compromise.
1 valid exploit after Mallory filtered fakes, detection scripts, and README-only repos.
Repository purpose: a minimal Python proof-of-concept exploit for CVE-2023-46229, demonstrating an SSRF condition in LangChain’s Embaas integration (`EmbaasBlobLoader`). Structure: - `README.md`: states the project targets CVE-2023-46229 and references an (absent in this archive) PDF report; warns for research use. - `poc_ssrf.py`: the only exploit code. - `LICENSE` (MIT) and `.gitignore`. Exploit flow (poc_ssrf.py): 1) Modifies `sys.path` to import LangChain from a hardcoded local directory. 2) Imports `EmbaasBlobLoader` and `Blob` from `langchain_community...` with a fallback to older `langchain...` paths. 3) Sets `embaas_api_url` to `http://127.0.0.1:9999` (loopback) and instantiates `EmbaasBlobLoader(embaas_api_url=..., embaas_api_key='test_key')`. 4) Creates a dummy blob (`Blob.from_data(b"hello")`) and calls `loader._get_documents(blob)` to force the loader to perform an outbound HTTP request (the script comments indicate a `requests.post` sink in the upstream loader implementation). 5) Treats a connection error referencing 127.0.0.1 as evidence the SSRF attempt occurred. Capabilities: - Outbound request redirection to arbitrary URLs supplied via `embaas_api_url` (demonstrated with loopback probing). This can be used for internal service/port probing and potentially accessing internal HTTP resources depending on the environment and any egress controls. No C2, persistence, or post-exploitation payloads are present; it is a focused SSRF reproduction script.
Products and vendors Mallory has correlated with this vulnerability. Open in Mallory to drill down to specific CPE configurations and version ranges.
Vendor-confirmed product mapping. Mallory continuously reconciles this list against your asset inventory.
4 sources tracked across advisories, community write-ups, and news. New activity surfaces here as Mallory finds it.
LangChain server-side request forgery vulnerability enabling access to internal services and potentially sensitive data.
A server-side request forgery vulnerability in LangChain's RecursiveUrlLoader caused by insufficient validation of externally supplied crawl target URLs.
A previously reported LangChain vulnerability related to prompt injection and code execution issues, mentioned as similar historical context only.
Query your assets running an affected version, and investigate the blast radius.
Every observed campaign linking this CVE to a named adversary.
Malware families riding this exploit, with evidence and IOCs.
YARA, Sigma, Snort, and vendor rules, auto-deployed to your SIEM.
Cross-references every affected SKU, including bundled OEM variants.
Community discussion across Reddit, Mastodon, and other social sources.