I understand that TLS handshake is done at the beginning of a session and is mainly about generating the symmetric key that is used to encrypt all later communications.
once the handshake is completed, then how often this handshake should happen (regenerate symmetric key) ?
where the symmetric key is stored on client side(browser) ?
how the server identifies the client on server side ?
Related
I have written separate client and server applications that use OpenSSL to exchange encrypted data. Both applications work. The only minor issue I have involves the client application. SSL_is_init_finished returns true after the client receives the first TLS handshake message. I need something that can be used to verify that the client has received the second (assumed to be last) TLS handshake message. Does the OpenSSL API provide a method?
I have traffic between clients (which send XML over HTTPS) to my IIS.
I tried to decrypt the traffic using Wireshark and the following settings:
Adding the private key:
But even after setting this, I can't see the decrypted data:
Should I change any other settings to see the original data?
There is strong possibility that a Diffie-Hellman (DH) key exchange is being used here. In that case Wireshark cannot decipher SSL/TLs with a private key. You can check for this in the handshake packet.
From the Docs:
The RSA private key file can only be used in the following
circumstances:
The cipher suite selected by the server is not using (EC)DHE.
The protocol version is SSLv3, (D)TLS 1.0-1.2. It does not work with TLS 1.3.
The private key matches the server certificate. It does not work with the client certificate, nor the Certificate Authority (CA)
certificate.
The session has not been resumed. The handshake must include the ClientKeyExchange handshake message.
My server may be attached to a RSA or ECDSA certificate.
In my code I set the prtocol method as
sslmethod = SSLv23_server_method();
and setting the cipher using the SSL_CTX_set_cipher_list() API.
The problem is when the server is attached to ECDSA certificate and if I pass "RSA" cipher SSL_CTX_set_cipher_list() does not return any error though "RSA" cipher is not compatible with ECDSA certificates.
Is there any openssl API which I can use to find out if the cipher we are going to set is compatible with the certificate or with protocol or not?
Is there any openssl API which I can use to find out if the cipher we are going to set is compatible with the certificate or with protocol or not?
If you author the app and own the server, then YES, you do know what to expect. But the general case is NO, you don't know in advance and you can't query for it. You try to setup a channel and if it fails, then you try something else.
Generally speaking, there's no way to know in advance what cipher suites a server supports; or what type of public key is bound in the server's certificate. In fact, there's no way to know what protocol versions (SSLv2, SSLv3, TLS 1.0, TLS 1.1, TLS 1.2, etc) a server supports. Its the reason for RFC 7405, TLS Fallback Signaling Cipher Suite Value (SCSV) for Preventing Protocol Downgrade Attacks.
In 2016/2017, about the best you can do is assume:
No SSLv2, SSLv3; only TLS 1.0 and above
Cipher suites from the collection "HIGH:!aNULL:!kRSA:!RC4:!MD5"
"!kRSA" is "no RSA key transport", and it means you can only use ephemeral key exchanges. If a server has a RSA key, then it will only be used for authentication. That is, the key will be used to sign the server's ephemeral values during key exchange so the client knows they are authentic.
What I want is test tls handshake when psk is active.
I also want to see every http header exchanged during the connection between client (my laptop) and public server.
Now I am wondering if there is a public psk tls server where I can do my test.
Regards.
I doubt that there is a web server using PSK on the internet open for public testing. Also I doubt that browsers support PSK cipher suites. But you can setup your own web server using PSK with openssl:
openssl s_server -psk 1a2b3c4d -nocert -www
And the matching client:
openssl s_client -connect 127.0.0.1:4433 -psk 1a2b3c4d
As for the HTTP protocol: it is independent from the TLS layer, i.e. it does not change if PSK or the normal authentication with certificates is used.
Even if some public TLS Server were to support PSK, you won't be able to test your client with it. There is a fundamental difference between the way public key authentication (which is used by most of the TLS Servers) work and PSK.
Public Key Authentication:
Incase of Public Key Server Authentication (the ones that doesn't involve Client Authentication), the server sends a Certificate, which contains a Public Key and Client encrypts it's pre-master secret and sends it to server which only the server can decrypt. In this way both have the same pre-master secret and can use the same set of derivations to further derive the final key.
Pre-Shared key:
As the name indicates the pre-shared requires both parties to have the same key pre-shared among themselves. They just exchange the IDs between them to indicate which of the Pre-Shared they will be using to generate the final key.
So, even if there is a server which supports PSK, you should have the same set of (or atleast one) of the keys which it has, which is impossible as those servers won't share their keys with anyone apart from whom it is supposed to be shared with (the legit clients).
So, the best way for you is to use openssl's test client and server tools and test it.
According to RFC client sends premaster secret encrypted with RSA public key of the server. But when they are using ECDHE as asymmetric algorithm, client key exchange message will contain the pulic key of the client. If this is the case ? when will the client send premaster secret and how ?
when will the client send premaster secret and how ?
It won't. See #8.1.2:
A conventional Diffie-Hellman computation is performed. The negotiated key(Z) is used as the pre_master_secret.