How do you determine if a REST webservice is using Basic, Kerberos, NTLM, or one of the many other authentication methods?
If it's a black box scenario, I usually connect with Fiddler, and inspect the actual traffic.
You send it a request, presumably get an HTTP 401 code, and look at the WWW-Authenticate
header that (per RFC 2616) the response MUST
include. If instead you get a 403 or some other weird status, or a missing WWW-Authenticate
header, you curse at website authors that don't follow the core HTTP RFC, and start sniffing the traffic to try to reverse engineer what nonstandard mess they've done this time;-).
When you send an unauthenticated request the service has to respond with a "HTTP/1.1 401 Unauthorized" and the response contains a WWW-Authenticate
header that specifies what authentication scheme is expected (Basic
, Digest
), the security realm and any other specific value (like Digets's nonce). So if the server responds with:
HTTP/1.0 401 Unauthorized
WWW-Authenticate: Digest realm="example.com",
qop="auth,auth-int",
nonce="...",
opaque="..."
it wants a Digest authentication. If the response looks like:
HTTP/1.0 401 Unauthorized
WWW-Authenticate: Basic realm="example.com"
then it wants a Basic authentication. Some (poorly) implemented servers/sites don't handle the Basic correctly and respond directly with 403 Forbidden instead of challenging first.
NTLM is similar in as the server reponds with a 401 and a WWW-Authenticate header with the value NTLM
, but there is no official public spec for it, since is Microsoft proprietary. There are various reverse engineered descriptions.
Unfortunately REST does not come with a WSDL style description of service to discover the authentication scheme used a priori.