NGINX ssl_client_certificate: "FAILED:unable to verify the first certificate" - ssl

On my NGINX server I'm using the ssl_client_certificate /path/to/clientcert.pem; option to set up a trusted client certificate.
This clientcert.pem file contains two trusted client certificates, the one (A) with the CA certificate of its Issuer, the other (B) one without it. So that's three certificates in total.
When client A makes are request everything works as expected, but when client B makes a request, even though the TLS handshake is successful, NGINX responds with 400 and html message including "The SSL certificate error". The $ssl_client_verify variable holds "FAILED:unable to verify the first certificate".
Is there a way to make NGINX accept the client certificate from B without knowing its Issuer? Or am I misunderstanding the problem here?
Edit: Feedback on #Seifeddine Besbes' comment asking openssl s_client output.
The relevant section is
---
Acceptable client certificate CA names
/C=ZA/ST=MyState/L=MyCity/O=Org/OU=OrgUnit/CN=example.com/emailAddress=CA#example.com
/C=ZA/ST=MyState/L=MyCity/O=Org/OU=OrgUnit/CN=clientA.example.com
/C=ZA/ST=MyState/L=MyCity/O=OtherOrg/OU=OrgUnit/CN=clientB.example.com
Server Temp Key: ECDH, X25519, 253 bits

Related

Using self-signed SSL certificate works, but CA signed certificate results in handshake alert failure 40 in response to client hello

So far, I've used openssl, sslyze, keystore utilities, and some of the standard Windows diagnostics commands to try and characterize this problem. The summary is that as soon as I attempt to use a keystore that has the CA signed certificate in it, I get an immediate handshake failure 40 after the client hello. The connection never gets as far as a server hello with the CA cert in place.
On the same machine (Windows Server 2012), if I use a self signed certificate, the connection works as expected.
Openssl shows the handshake failure with the CA certificate in place.
sslyze shows all ciphers rejected with either no ciphers available or TLS / Alert handshake failure.
The fact that this is successful with a self-signed certificate tends to indicate that the "basics" are there, that the client and server have the necessary ciphers and so on to connect, and that server.xml is configured properly.
There were no error messages when importing the certificates, and everything seems to be configured correctly with the CA certs based on keytool checks. There are no error messages given during server start to indicate any problems processing the certificate.
Is there something I should be looking more closely at in regards to the keystore with the CA cert that could cause this complete rejection of the client hello? How would a different keystore or CA certs within aaffect the earliest steps of the handshake?
Thank you for information provided.
I assume that you are talking about a server side certificate and a server side key store, since the error happens at a state where no client certificates are yet involved. If the server is able to send the self-signed certificate to the client, but is not able to send the CA signed certificate to the client, then it must be something wrong with the certificate you are trying to send or that the certificate can not be used together with the ciphers offered by the client.
Since the problem is obviously on the server side you should first check all logs written on the server side for hints what the error might be. Typical problems are non-existing files, wrong files, password protected client key w/o providing a password or that the key does not belong to the certificate.

Browser doesn't apply client certificate: 403.7

I'm trying to set up client certificate authentication. I was able to generate a CA-, server- and client-certificate. As long as I use Fiddler everything works as expected. However, as soon as I start using a browser it doesn't work anymore (HTTP Error 403.7 - Forbidden).
Of course I imported the client certificate in the Personal store and I made sure Client Certificate Negotiation is enabled.
I also tried openssl s_client -connect 127.0.0.1:443 -state -debug but I couldn't really make sense of the result... The only thing what's weird is that my CA doesn't show up in the Acceptable client certificate CA names section.
Anything else I could try?
Update:
I think it doesn't matter but my server certificate is set up for 127.0.0.1. Therefore I'm using https://127.0.0.1/... in my browsers.
Update2:
Using Wireshark I noticed that my servers' response depends on the client:
Fiddler (OK):
Client Hello
Server Hello, Certificate, Server Hello Done
Browser (Not OK):
Client Hello
Server Hello, Change Cipher Spec, Encrypted Handshake Message
Update3:
After enabling clientcertnegotiation the server response is different but still doesn't work:
Server Hello, Certificate
Certificate Request
Certificate, Client Key Exchange, Change Cipher Spec, Encrypted Handshake Message
My self-signed CA doesn't seem to be in the Distinguished Names list...
Update4:
SSL Settings: Checked Require SSL and Client certificates set as Required. Client cert shows up in Personal and the intended purpose is Client Authentication.
I finally found the issue and a workaround:
As mentioned in Update3, Distinguished Names doesn't contain my CA. This is because Distinguished Names has a limit of 2^14 bytes (16384 bytes). Because I do have a lot of CA installed on my machine my CA simply didn't make it in. The TLS standard would allow to send multiple messages but unfortunately Windows doesn't support this!
As mentioned here you have a few possibilities. The simplest one is this:
At your server add a DWORD (not QWORD!) value called SendTrustedIssuerList in your registry under HKEY_LOCAL_MACHINE\SYSTEM\CurrentControlSet\Control\SecurityProviders\SCHANNEL and set it to 0. This will prevent your server from sending a list at all, letting the client choose from any installed client certificate.
Unfortunately I couldn't see any traces in the Event Viewer (as reported elsewhere). Therefore the issue wasn't easy to spot (I had to use Wireshark in order to check Distinguished Names).
Use the Accept option instead of the Require option of the "Client certificates" feature.
In IIS Manager, locate the Web application for which you want to change the SSL setting.
In Features View, double-click SSL Settings.
On the SSL Settings page, select the Accept option under Client certificates.
In the Actions pane, click Apply.
More info here
Client certificate should be imported in CurrentUser\My store with private key (i.e. p12 or pfx file usually).
CA certificate should be in LocalMachine\Root store so that IIS trusts all certificates issued by the CA and the CA is trusted for every user on the computer.
CRL issued by the CA should be either available through URL (specified in every end entity certificate that CA issued) or imported in LocalMachine\My store.
NOTE: openssl doesn't use windows certificate store so this will have no efect on openssl s_client -connect 127.0.0.1:443 -state

SSL client verification failure

I have a backend server with SSL only as defaut public interface. This server listens on 443 with cert, key and ca files on, ssl verify client is set to true.
On client side, I have client cert, key and a ca file made of 2 CA files catted together (intermediate CA, root CA) in a single ca.pem file. When I do curl request on my backend server, it works fine. I also tested the server / client cert and key with gnutls-serv and openssl s_server to ensure all was valid.
But when I create a keystore on Apigee (client cert and key file [pem format]) and a trustore (ca.pem file), there is an error:
- target.name fsbca-test
- Properties
- Expression ("fsbca-test" equals target.name)
- ExpressionResult true
- Tree TARGET_fsbca-test
- error The Service is temporarily unavailable
- error.cause General SSLEngine problem
- error.cause.cause General SSLEngine problem
- error.class com.apigee.messaging.adaptors.http.HttpAdaptorException
- state TARGET_REQ_FLOW
- type ErrorPoint
If I put SSL verify client to false on my backend, then the request is correctly diverted by Apigee and I get the response.
If I put IgnoreValidationErrors to true in the target endpoints property, then the request is diverted to my backend server but I can see an error in the server's log: "client sent no required SSL certificate while reading c...".
Any ideas about what could be wrong in what I am doing?
Additional track: could it be an issue on Apigee side with the CA file made of 2 certs (it may ignore trailing certificates found in a .pem). If wanted to test pkcs12 and jks but I failed to upload them to Apigee (the API doc page only describes .pem, JAR and cert action). I wrote a small Java client with pkcs12 keystore and jks trustore and it worked fine from my local workstation.
Thank you in advance for any piece of information that could help me.
Regards
Fr
You should upload your certificates separately (one per cert), and you need the entire trust chain of certificates to be stored in your truststore.
Here is the page about SSL to your backend.
EDIT:
Here is a method I know works:
1) Separate certs into separate PEM files. Do not put more than once cert in a file.
2) Validate each cert using openssl:
openssl x509 -noout -text -in <cert file name>
Validate that no certs are expired, and that the Subject and Issuer fields create a chain of all the certs, with identical names.
The server's certificate, the root certificate, and all certificates in between need to be in the truststore.

How to determine a server's list of CA certificates that it will accept from client?

According to https://wiki.jasig.org/display/CASUM/X.509+Certificates,
After the Server sends the certificate that identifies itself, it then can then send a list of names of Certificate Authorities from which it is willing to accept certificates.
I am wondering how to determine what this list is, and how to modify it.
The reason I am asking is that I am getting an infinite redirect between my server and my client after successful validation (i.e., the ticket stage), and I think it has to do with the CAS server not recognizing the CAS client's certificate (the client's certificate is self-signed).
If you want to see what this list is, you can use OpenSSL:
echo "" | openssl s_client -connect your.server:port
This will show various messages regarding the handshake, including the certificates and the list of CAs in the CertificateRequest message.
Ultimately, it's determined by the active X509TrustManager's getAcceptedIssuers() method. By default, this will be the list of Subject DNs of all your trust anchors (that is, the Subject DNs of all the certificates in your trust store).
Your client certificate will have to be verified by the server. This is normally done during the handshake by the trust manager, which (unless tweaked) will build a chain to a known CA (or at least known cert if it's the user cert itself) in the trust store.
Adding your self-signed certificate to your trust store should be sufficient. It doesn't have to be the cacerts file bundled with the JVM, you could make a copy of it and use the trust store settings of Apache Tomcat's connector to set it up.

OpenSSL: unable to verify the first certificate for Experian URL

I am trying to verify an SSL connection to Experian in Ubuntu 10.10 with OpenSSL client.
openssl s_client -CApath /etc/ssl/certs/ -connect dm1.experian.com:443
The problem is that the connection closes with a Verify return code: 21 (unable to verify the first certificate).
I've checked the certificate list, and the Certificate used to sign Experian (VeriSign Class 3 Secure Server CA - G3) is included in the list.
/etc/ssl/certs/ca-certificates.crt
Yet I don't know why it is not able to verify the first certificate.
The entire response could be seen here:
https://gist.github.com/1248790
The first error message is telling you more about the problem:
verify error:num=20:unable to get local issuer certificate
The issuing certificate authority of the end entity server certificate is
VeriSign Class 3 Secure Server CA - G3
Look closely in your CA file - you will not find this certificate since it is an intermediary CA - what you found was a similar-named G3 Public Primary CA of VeriSign.
But why does the other connection succeed, but this one doesn't? The problem is a misconfiguration of the servers (see for yourself using the -debug option). The "good" server sends the entire certificate chain during the handshake, therefore providing you with the necessary intermediate certificates.
But the server that is failing sends you only the end entity certificate, and OpenSSL is not capable of downloading the missing intermediate certificate "on the fly" (which would be possible by interpreting the Authority Information Access extension). Therefore your attempt fails using s_client but it would succeed nevertheless if you browse to the same URL using e.g. FireFox (which does support the "certificate discovery" feature).
Your options to solve the problem are either fixing this on the server side by making the server send the entire chain, too, or by passing the missing intermediate certificate to OpenSSL as a client-side parameter.
Adding additional information to emboss's answer.
To put it simply, there is an incorrect cert in your certificate chain.
For example, your certificate authority will have most likely given you 3 files.
your_domain_name.crt
DigiCertCA.crt # (Or whatever the name of your certificate authority is)
TrustedRoot.crt
You most likely combined all of these files into one bundle.
-----BEGIN CERTIFICATE-----
(Your Primary SSL certificate: your_domain_name.crt)
-----END CERTIFICATE-----
-----BEGIN CERTIFICATE-----
(Your Intermediate certificate: DigiCertCA.crt)
-----END CERTIFICATE-----
-----BEGIN CERTIFICATE-----
(Your Root certificate: TrustedRoot.crt)
-----END CERTIFICATE-----
If you create the bundle, but use an old, or an incorrect version of your Intermediate Cert (DigiCertCA.crt in my example), you will get the exact symptoms you are describing.
SSL connections appear to work from browser
SSL connections fail from other clients
Curl fails with error: "curl: (60) SSL certificate : unable to get local issuer certificate"
openssl s_client -connect gives error "verify error:num=20:unable to get local issuer certificate"
Redownload all certs from your certificate authority and make a fresh bundle.
I came across the same issue installing my signed certificate on an Amazon Elastic Load Balancer instance.
All seemed find via a browser (Chrome) but accessing the site via my java client produced the exception javax.net.ssl.SSLPeerUnverifiedException
What I had not done was provide a "certificate chain" file when installing my certificate on my ELB instance (see https://serverfault.com/questions/419432/install-ssl-on-amazon-elastic-load-balancer-with-godaddy-wildcard-certificate)
We were only sent our signed public key from the signing authority so I had to create my own certificate chain file. Using my browser's certificate viewer panel I exported each certificate in the signing chain. (The order of the certificate chain in important, see https://forums.aws.amazon.com/message.jspa?messageID=222086)
Here is what you can do:-
Exim SSL certificates
By default, the /etc/exim.conf will use the cert/key files:
/etc/exim.cert
/etc/exim.key
so if you're wondering where to set your files, that's where.
They're controlled by the exim.conf's options:
tls_certificate = /etc/exim.cert
tls_privatekey = /etc/exim.key
Intermediate Certificates
If you have a CA Root certificate (ca bundle, chain, etc.) you'll add the contents of your CA into the exim.cert, after your actual certificate.
Probably a good idea to make sure you have a copy of everything elsewhere in case you make an error.
Dovecot and ProFtpd should also read it correctly, so dovecot no longer needs the ssl_ca option.
So for both cases, there is no need to make any changes to either the exim.conf or dovecot.conf(/etc/dovecot/conf/ssl.conf)
If you are using MacOS use:
sudo cp /usr/local/etc/openssl/cert.pem /etc/ssl/certs
after this Trust anchor not found error disappears
For those using zerossl.com certificates, drag and drop all certificates (as is) to their respective folders.
Cut and pasting text into existing files, may cause utf8 issues - depending upon OS, format and character spacings.