Mutual Authentication TLS - authentication

What is the use of mutual authentication in TLS without restricting the client cert?
Here is my understanding about client/mutual auth using TLS.
The idea is that both the server the client authenticate/verifies each other certs so,
1- The client verifies the server cert based on its CA trust store
2- The server verifies the client cert based on its *CA trust store*
Now the key point to me is the second one, the server has to trust the client certificate by either saving it in the server trust store, or save the CA/ICA of the client cert, which should be private to the client, not via public CA, private CA to that client that the server wishes to trust.
Now [rfc5246][1] says the following
If the client has sent a certificate with signing ability, a digitally-signed
CertificateVerify message is sent to explicitly verify possession of
the private key in the certificate.
This won't achieve any authentication correct? for example, if I have a server that trusts any certificate signed by the trusted CAs around the world, then why bother at all with client authentication at all?
[1]: https://www.rfc-editor.org/rfc/rfc5246

When the server gets the client cert (as part of the TLS protocol), the server does all the normal cert checks, including chaining up to a trusted root. For the server to trust a client cert issued by Foo CA, the server needs to have the Foo CA root already installed at the server.
The corner stone of X.509 certs is root CA certs. A host should only trust certs issued by the CAs it trusts. So when an admin installs FooCA's roots, the admin is stating "I trust the certs issued by Foo and I trust that Foo did appropriate checks that the cert is valid and assigned to the correct entity."
The server does not need to store the client's cert, there's no reason to. When the cert comes in as part of TLS, the server simply checks it. No persistence needed, and if anything fails, and that includes not having the Foo CA root cert installed, then the connection fails.
The server DOES authenticate the client. A certificate binds a public key (in the cert) to an entity; for example CN=frodo#theshire.com. When the client connects and the server asks for the client cert, the server will check that the name in the cert (frodo#theshire.com) matches the name of the user, but it will also get the client to encrypt some data using the private key associated with the public key in the cert, and if the server successfully decrypts the data, then the client (frodo#theshire.com) really is who they claim to be. Or, in the worst case, an imposter who got access to Frodo's private key :)
Does that help?

Related

Which client certificate auth validations are done by NGINX for auth-tls-verify-client = optional_no_ca

We have a K8S service leveraging NGINX and in some flows would like to accept client certificate authentication.
Service has a dynamic list of public trusted client certificates (PEM format), and the root CAs aren't known.
In NGINX, it seems like the best setting to use would be:
nginx.ingress.kubernetes.io/auth-tls-verify-client: optional_no_ca
While sending the full certificate ($ssl_client_escaped_cert) to the upstream service to compare the entire public cert.
The question is whether NGINX will still perform the client cert validations during SSL handshake (and only skips CA checks), to verify the request is indeed sent by the one and only owner of the cert and its private key.
It will still check in the TLS handshake that the public key in the certificate can be used to verify the signature in CertificateVerify, i.e. that the client actually owns the private key to the sent certificate.
It will not check that the certificate itself is issued by a trusted CA etc - such verification are expected to be done elsewhere.

Does TLS client ever require a certificate installation?

In a TLS communication I always thought the server would send the public certificate to the client during the handshake process.
However, I received a request from a provider asking us to manually install a certificate in order to initiate a TLS communication with one of their STunnel server.
I'm a bit confuse by this request. Is there's a situation where TLS server would not be able to send the certificate to the client and therefore required a manuel certificate installation?
There are two cases where a certificate installation is needed on the client side:
The server is using a certificate not issued by a public CA, i.e. self-signed or issued by a private CA. In this case the client must import the certificate or CA up front and install it as trusted.
The server is requiring client authentication based on certificates, i.e. not only the server authenticates to the client with a certificate but also the client to the server (instead or additionally to typical passwords). In this case the client has to get a client certificate and matching key and import so that it can be used for authentication.
It is unclear which of these cases is the one you are dealing with.

SSL/TLS handshake between Websphere MQ Server and Client

I'm debugging SSL errors between Websphere MQ Server and Client using T.Rob's suggestions and need help understanding the SSL handshake (SSL connect to MQ using .net mq client SSLV3?).
My WMQ 7.5 client application is C code and uses a keystore (.kdb). Utilizing a CHLTAB provided by the WebSphere admin. The WMQ server is running Java and the channel is defined with mutual authentication.
The article states that in the SSL/TLS handshake, the server always sends its public certificate in response to a connections request. The client then must validate that certificate by first checking the signature and validity date, then looking in its trust store for the thing that signed the certificate.
Here's my confusion: Since my keystore on the client side has only the application personal cert, how does the client validate the public certificate sent by the server? I have provided the Common Name of my application cert to the WebSphere server admin but nothing more than that.
Thanks in advance for the clarification!
The bit about "my keystore on the client side has only the application personal cert" is troubling. That won't work. The client KDB must have the server's public key. If the MQ server has SSLCAUTH(OPTIONAL) the the server's public cert is all that's required in the KDB for the connection to succeed.
The first part of the TLS handshake is where the client validates the server's certificate. The use of public/private key pairs is how the authenticity of the thing on the other side is assured. In order for that to happen the server must have its own personal certificate and the client must have the public key of the root of the signer chain. In the case of self-signed certificates the public portion of the personal cert must be trusted by the client. In the case of CA-signed certs the CA Root must be trusted by the client. Whichever one it is, the client must trust something used to sign the server's personal cert or that cert can't be validated.
The TLS handshake is symmetrical so the second part works exactly the same as the first but the roles are reversed. Therefore, in cases where mutual authentication is enabled the client must have its own personal cert (because that contains the private key) and the server must trust whatever signed the client's matching public key. If the client cert is self-signed the QMgr has to trust it. If the client's cert is CA-signed, the QMgr must trust the signer. Either way, the QMgr has to have a cert to validate the client in its KDB.
Following this logic, for anonymous client connections the required parts are the personal certificate in the QMgr's keystore (because it contains the QMgr's private key), and the matching trusted certificate in the client's KDB or for Java the Trust Store. Neither of these is optional.
If the client is to be authenticated you still need the same two certs as for anonymous clients because that part of the handshake has to complete before the client is authenticated. In addition, now you also need the client to have its own personal certificate (because it contains the client's private key) and the QMgr now needs to trust whatever signed the client's certificate - the client cert if self-signed or the signer root if CA-signed.
As a side note, There is also some confusion in the post because it says that "My WMQ 7.5 client application is C code and the WMQ server is running Java." There is nothing in the queue manager that uses Java on the server side. There are Java components installed to do things like manage JNDI objects and run the sample code. In modern MQ versions Java runs the Web Console. But there are no Java components in the QMgr itself and no Java components in the path of the incoming channel connection request. That's all habdled by the QMgr's listener, agent, and other internal processes. So I'm not at all sure what is being referred to there other than that the notion of Java running on the MQ server side and participating in the TLS handshake might be a source of some of the confusion mentioned in the post. ;-)

SSL approach for private software

What is the proper way of using SSL certificates for private applications? By private I mean that I am the only user, and software is running on my computers.
I want to have a encrypted communication between two of my programs. I want to send passwords between them, so I need to be sure that remote program is not fake/hacked.
As far as I understand I don't need to get paid SSL certificate from the CA, if there is no third party involved.
Is the following correct?
Server has a private key and self-signed SSL certificate.
Client has a copy of server's self-signed certificate (it needs to be well protected).
During the handshake server sends the certificate to client.
client checks if the certificates are the same.
client can start encrypted transmission.
Is there other way?
Server has a private key and self-signed SSL certificate.
Yes
Client has a copy of server's self-signed certificate (it needs to be well protected).
The client has either a copy of the certificate or the certificates public key or the fingerprint of these. Since the certificate is public these information do not need to be protected. Only the private key of the server (residing only in the server side) needs to be protected because using this key one could prove ownership of the certificate.
During the handshake server sends the certificate to client.
Yes.
client checks if the certificates are the same.
Kind of. It might check the certificate or the public key or the fingerprints.
client can start encrypted transmission.
Yes.
I would recommend that you read the OWASP article about certificate and public key pinning. It also contains sample code for various environments.
Client has a copy of server's self-signed certificate (it needs to be well protected).
Clients do not have copy of the server certificate. They get it in SSL handshake
client checks if the certificates are the same.
NO! Clients will have the public certificate of the Certificate Authorities who would have signed the server certificate. They will validate the server cert with the CA cert including things like certificate expiry, CRLs. Not compare for 'sameness'
In your case you are using the self-signed certificates. The clients should be made to ignore the self signed certificate and proceed with SSL handshake.
I would recommend you read through SSL handshake sequence again.

SSL and public key security

I'm consuming a web service from an Android device using HTTP(s) with SSL. A self-signed (untrusted) certificate is used for client authentication.
I have a general understanding of how public/private keys are used for SSL. From my understanding I can clearly see how a certificate can be used to set up a secure connection and transmit data securely. However I do not understand how they are used for client authentication since the certificate contains the public key and is not kept a secret.
I have a few questions:
Where can I read about how SSL and certificates can be used for client authentication?
Even if the certificate was not made public...by visiting the HTTPS URL in a browser I can view and save the certificate. I can then package up the certificate in a key store and use it from an application.
In this post Jeremy Huiskamp writes
client auth will automatically be performed when the server requests
it
...so client authentication as well as encryption of data can be performed with certificates?
Edited to answer the first part of my question: The client keystore should contain not only the server's public key but also the client's private key. The server must then be able to decrypt using the client's public key? Does this mean the keystore should have two certificates?
First, a quick point about the terminology in public key cryptography:
you sign and decrypt/decipher using a private key,
you verify (a signature) and encrypt/encipher using a public key.
(You don't really "decrypt" using a public key.)
Using SSL/TLS with or without client-authentication, the server presents a certificate (*) for which it has the private key. The server sends its certificate during the SSL/TLS handshake (at the beginning of the connection) and is able to decipher what the client sends using its private key (which it keeps private). The private key and certificates are stored in the server's keystore (or equivalent if it's not implemented in Java).
As part of this, the client uses its truststore, which is a form a keystore that contains trusted certificates, to verify the server certificate. The server certificate could be trusted by being explicitly in the truststore or, in most cases, trusted by linking in to a trusted CA certificate in the truststore (PKI).
The terminology between keystore and truststore in Java can be a bit confusing, you can find more details in this answer.
Regarding your question, the client's truststore doesn't contain the server's public key, but either its certificate or a CA certificate with which it should be verifiable. (It's not just about having the public key, but knowing whose it is, using the other pieces of information in the certificate.)
When you use client-certificate authentication in addition to this, there is a truststore (or equivalent) on the server side and a keystore on the client side too, since the roles are reversed for this purpose.
In the SSL/TLS handshake that uses client-authentication, the server requests a certificate from the client, who sends it (if available).
At the end of this handshake, the client sends a CertificateVerify message, which signs all the messages exchanged so far between the client and the server (so it's something known to both) using the client certificate private key. The server is then able to verify this signature against the public key within the client certificate it has obtained as part of this exchange. This proves to the server that whoever is on the client side has the private key corresponding to the public key in the certificate it has sent.
The next step for the server is to verify whether to trust this certificate, i.e. whether to trust the binding between identity and public key as presented and "sealed" within the certificate.
This is usually done using a PKI, whereby you check the certificate against a known CA, or if your deployment environment is sufficiently small, against a fixed set of trusted certificates. (There can be alternative methods of verification, but their usability will really depend on the circumstances where you want to deploy this system.)
Therefore, for your second question:
The client keystore should contain at least the client's certificate and its private key.
The client truststore should contain the server certificate or a CA certificate with which the server certificate can be verified.
Since both keystore and truststore are a keystore (in the storage format sense, usually a file) used for a different purpose, it's often possible to use the same keystore to serve both the purpose of keystore and truststore.
(*) There are cipher suites that don't rely on certificates, but that's unusual and off topic for this question.
A certificate just binds an identity to a public key. This binding is not secret, so there is no need to keep the certificate secret. If I have John Smith's certificate, I can prove that John Smith owns the secret key corresponding to a particular public key. But since I don't know that secret key, the certificate is of no use to me.
When authentication occurs by certificate, one step is always to have whoever presents the certificate prove they know the secret key corresponding to the public key in the certificate. If you cannot pass that step, the authentication fails.
The server's keystore should have the server's certificate. The client's keystore should have the client's certificate. The client will present its certificate to the server, so the server will learn the client's public key that way. (And vice-versa.)