Is port 9200 (Elasticsearch) exposed to the internet without authentication?
Port 9200 is Elasticsearch's default HTTP port, the interface for search and index requests. Before version 8.0, Elasticsearch's security features were opt-in and disabled by default, so a new node would answer unauthenticated requests to read, modify or delete any index. 9200 isn't on AdminKit's port-check tool's nine-port list, so this guide explains how to confirm and secure your exposure.
By AdminKit team, Netbase JSC · Updated
Key takeaways
- Port 9200 is Elasticsearch’s REST API, and every search, write and administrative call goes through it. Port 9300 carries node-to-node traffic.
- Security (users, roles, TLS) became free in 6.8.0 and 7.1.0 and switched on by default in 8.0. Clusters started on older versions, or with
xpack.security.enabled: false, answer anyone. - On 8.x the automatic setup adds
http.host: 0.0.0.0, so the HTTP port listens on every interface by default, protected by TLS and a password rather than by the bind address. - From outside,
curltells you the state: cluster JSON means no authentication,401means security is on but the API is reachable. - Fix it by turning security on, binding HTTP to a private address, firewalling 9200 and 9300, and upgrading anything still on a release with scripting flaws in CISA’s exploited list.
Why was a no-password cluster the default for so long?
Elasticsearch’s own networking settings documentation confirms 9200 as the default port for the Elasticsearch HTTP API that every search, index and administrative request flows through. For years, the security features that gate that API behind authentication — user accounts, role-based access control, and TLS — were part of a paid tier and disabled by default; a cluster started with no extra configuration simply answered any request that reached it.
That changed in two steps. In May 2019 Elastic made its core security features free in versions 6.8.0 and 7.1.0: TLS, the file and native user realms, and role-based access control. They were still off unless an administrator configured them. Elastic’s own 8.0 release notes confirm the second step: as of version 8.0, security is enabled by default for self-managed clusters, with authentication, role-based access control, and encrypted communication switched on automatically rather than left for an administrator to opt into.
| Version range | Security out of the box | What an exposed node gives away |
|---|---|---|
| Before 6.8.0 / 7.1.0 | Not in the free distribution | Everything, to anyone who reaches 9200 |
| 6.8.x and 7.x | Free, but off until configured | Everything, unless someone enabled it |
| 8.0 and later | On: TLS, elastic password, roles | A login prompt, unless security was turned off |
Which ports does Elasticsearch actually open?
Two listeners matter. The networking reference gives http.port a default range of 9200-9300, taking the first free port, and transport.port a default range of 9300-9400 for traffic between nodes. In practice a single node lands on 9200 and 9300, and a second node on the same host lands on 9201 and 9301. IANA has no Elasticsearch entry at either number; its registry lists 9200 as wap-wsp, an old WAP Forum service.
The bind address is where version 8 surprises people. network.host defaults to _local_, the loopback addresses, but the automatic security setup that runs on a new 8.x node’s first start generates TLS certificates for HTTP and transport, sets a password for the elastic superuser, and adds http.host: 0.0.0.0 to elasticsearch.yml, which overrides the loopback default for HTTP. A fresh 8.x node therefore listens on every interface for 9200, and its protection is the TLS and password layer. Turning security off on such a node removes the only barrier between the API and any network that can route to it.
The transport side has a guard of its own. Elastic’s bootstrap checks put a node into production mode once transport binds to a non-loopback address, and in that mode a failed check stops Elasticsearch from starting. Those checks cover memory, file descriptors and similar system settings, not whether authentication is on, so passing them says nothing about exposure.
What keeps going wrong anyway?
The problem now is mostly inherited exposure: clusters deployed on a pre-8.0 version that were never upgraded, or that had security manually disabled during setup “to get it working” and never turned back on. Researchers keep finding the result. SOCRadar reported finding a publicly reachable Elasticsearch server, requiring no authentication at all, that held approximately 544 million records in URL, email and plaintext-password format — one entry in a long-running pattern of researchers discovering large, completely open Elasticsearch indices simply by connecting to port 9200 with no credential whatsoever.
What does an exposed cluster actually allow?
An unauthenticated Elasticsearch REST API on port 9200 typically allows anyone to list every index on the cluster, read any document in any of them, and — depending on the specific API calls permitted — write new documents or delete indices outright. There’s no concept of a read-only guest account at that point; a connection either has no access (behind authentication) or has essentially the same access as an administrator (with security disabled).
| Request | Effect on an unprotected cluster |
|---|---|
GET /_cat/indices?v | Lists every index with document counts and sizes |
GET /<index>/_search | Returns the documents in an index, page by page |
DELETE /<index> | Deletes the index and its data |
PUT /_cluster/settings | Changes cluster-wide settings |
Old releases add code execution on top. CVE-2014-3120 covers Elasticsearch before 1.2, whose default configuration enabled dynamic scripting, so a search request could carry MVEL expressions and Java code. CVE-2015-1427 covers a Groovy sandbox bypass in versions before 1.3.8 and 1.4.x before 1.4.3 that let a crafted script run shell commands. CISA added both to its Known Exploited Vulnerabilities Catalog in March 2022. Those versions are long out of support, which is exactly why a forgotten node still running one is worth finding.
How do you check your own exposure?
AdminKit’s open ports check checks a fixed nine-port list, and port 9200 isn’t on it — confirm your public IP with what is my IP, then test the port manually from a machine outside your network with nc -vz -w 3 203.0.113.5 9200. Confirming this directly matters because “we’re on a recent version” doesn’t guarantee security is actually turned on if it was ever manually disabled along the way.
The AdminKit result covers only the public address of the browser that runs it, so it describes your own office or home connection. If that result shows 80, 443 or 8080 open, check whether one of those forwards leads to a reverse proxy in front of Elasticsearch, since a proxy rule can expose the API without 9200 itself being open.
Because the API speaks HTTP, curl shows which state the cluster is in. Run it against your own address only:
$ curl -s -m 5 http://203.0.113.5:9200/
$ curl -sk -m 5 -i https://203.0.113.5:9200/
Example replies (illustrative; documentation address, exact text varies by version):
| Reply | What it means |
|---|---|
JSON with cluster_name, version.number and "tagline" : "You Know, for Search" | No authentication. Anyone can read and delete data. Close the port first. |
HTTP/1.1 401 Unauthorized with WWW-Authenticate: Basic realm="security" | Security is on. The API is still exposed to password guessing and future flaws. |
Empty reply over http://, but 401 over https:// | An 8.x node with HTTP TLS on, as the automatic setup configures it. |
| Timeout or connection refused on both | Nothing reachable on 9200 from where you tested. |
How do you confirm the listener on your own host?
On the node itself, look at the sockets and at the settings the running node reports:
$ ss -ltnp | grep -E ':(9200|9300) '
LISTEN 0 4096 *:9200 *:* users:(("java",pid=1207,fd=402))
LISTEN 0 4096 [::1]:9300 [::]:* users:(("java",pid=1207,fd=398))
GET _nodes/_local/http?filter_path=nodes.*.http.bound_address
GET _nodes/_local/settings?filter_path=nodes.*.settings.xpack.security.enabled
The ss lines are an example of the 8.x pattern: HTTP on every interface, transport on loopback. The two API calls return the addresses the HTTP layer bound to and whether xpack.security.enabled was set explicitly in the node’s settings. On a node where that setting is absent, check elasticsearch.yml for the block the automatic setup wrote.
How do you fix an exposed Elasticsearch cluster?
- If you’re on Elasticsearch 8.0 or later, make sure
xpack.security.enabledwas never turned off — Elastic ships it on by default, but a manual override during setup can disable it. - If you’re on an older version, upgrade and explicitly enable the security features rather than relying on network isolation alone.
- Bind Elasticsearch’s HTTP interface to a private network address instead of a public one, and put a firewall or authenticating reverse proxy in front of port 9200 so it’s never directly reachable from the internet.
- Firewall 9300 as well; only other nodes in the cluster need it.
- Audit existing indices for data that shouldn’t have been reachable, since researchers and automated scanners alike continuously probe for open 9200 instances.
A minimal elasticsearch.yml for a node that serves applications on one private network keeps the security block the automatic setup writes and narrows the bind address:
network.host: 10.0.2.21
http.port: 9200
xpack.security.enabled: true
xpack.security.http.ssl.enabled: true
xpack.security.transport.ssl.enabled: true
The certificate paths that go with the two ssl lines are the ones the automatic setup generated on first start. Exposed search clusters follow the same pattern as exposed document databases — see is port 27017 (MongoDB) exposed, and open Redis and other unauthenticated databases for the full outside-in walkthrough across data stores.
Which servers may you test with these checks?
Every check this page links to is for assets you own or are explicitly authorized to test. Never scan, probe, or attempt to connect to a network or device you don’t control — AdminKit’s active checks only ever run against the IP making the request or an asset you’ve verified ownership of. See our Disclaimer and Acceptable Use Policy for the full terms.
Elasticsearch’s HTTP API is one entry in AdminKit’s broader machine protocols cluster, which groups the network services and data-store protocols that routinely turn up reachable from the open internet with no authentication in front of them.
This guide is general information, not a security audit of your specific network — see our disclaimer for what AdminKit's guidance does and does not commit to.
Frequently asked questions
Does AdminKit's port-check tool scan port 9200?
No. The tool checks a fixed list of nine specific TCP ports (23, 554, 1883, 8883, 502, 7547, 80, 443, 8080) chosen for how often they turn up misconfigured on home and small-office devices — see the port-check tool page for the current list. 9200 isn't on that list, so confirm your own exposure with the manual check below.
Is 9200 an officially assigned port for Elasticsearch?
No — IANA's registry lists 9200 as wap-wsp, a WAP Forum service, not Elasticsearch; 9200 is simply Elastic's own documented default for the HTTP API, configurable via the http.port setting. That matters because an open 9200 doesn't guarantee Elasticsearch is behind it, so confirm what answers before drawing conclusions.
Didn't Elastic fix the no-authentication-by-default problem?
Yes, as of version 8.0 — Elastic's own release notes confirm security features including authentication, role-based access control and encrypted communication are enabled by default for self-managed clusters. The risk now is almost entirely clusters that were deployed on an older version and never upgraded or had security manually re-enabled, or that disabled it for convenience during setup and never turned it back on.
How bad is it in practice if my cluster is actually exposed?
Severe — an unauthenticated Elasticsearch REST API lets anyone list every index, read every document, and in many cases write or delete data outright. In one case SOCRadar reported finding roughly 544 million plaintext credential records in a single open, unauthenticated Elasticsearch server.
What's the fastest fix if I find my own cluster's port 9200 open?
If you're on 8.0 or later and security was manually disabled, re-enable xpack.security.enabled; if you're on an older version, upgrade and enable the security features explicitly, then bind Elasticsearch to a private interface and put a firewall or reverse proxy in front of it rather than exposing 9200 directly.
Sources
- Elastic — Networking settings (default http.port, network.host binding)
- Elastic — What's new in Elastic 8.0 (security enabled by default)
- Elastic — Security for Elasticsearch is now free (6.8.0 and 7.1.0, May 2019)
- Elastic Docs — Automatic security setup for self-managed clusters
- Elastic Docs — Bootstrap checks (development vs production mode)
- NVD — CVE-2014-3120 (Elasticsearch dynamic scripting)
- NVD — CVE-2015-1427 (Elasticsearch Groovy sandbox bypass)
- CISA — Known Exploited Vulnerabilities Catalog
- IANA Service Name and Transport Protocol Port Number Registry
- SOCRadar — Critical Elasticsearch Exposure: 544M Plain-Text Credentials Found Publicly Accessible
Want AdminKit to watch this for you?
Free checks tell you the state today. Protective DNS and monitoring tell you when it changes.
Talk to an expert
Huy Nguyen (David), Founder & CEO, Netbase JSC
Ask whether this fix applies to your setup, or get a second pair of eyes on your DNS and mail records.