I'm working on designing a server, in which the protocol allows for the client to open additional physical connections to the server which operate in the context of a single logical connection.
One thought I had was that, if we're using ssl, we'll need to do another ssl handshake for the new connection. It seems to me that it should be possible to send some secret to the client over the original, secure connection that would allow the new connection to be securely established without a handshake (similarly to what I've read about ssl session reuse).
Is this actually possible?
SSL already does this. Provided both ends support it, there is a feature called 'session resumption' which allows a new connection via an existing SSL session, via a much abbreviated handshake, without the certificate exchange and negotiation of shared secrets.
Yes, by reusing SSL Session. You can do this by using PEM_write_bio_SSL_SESSION and PEM_read_bio_SSL_SESSION and then adding it to SSL Context.
Keep in mind you only need to do this in client mode, server mode does this automatically for you.
Related
I am developping a FTPS client in a Embedded system with LWIP and mbedTLS stacks.
For now, what I do is :
Get an IP adress with DHCP
Get IP address of the server with DNS
Start a TCP connection to the server ( the control connection)
Telling the FTP server that I want to work with the TLS protocol with the command AUTH TLS
Handshake of the TLS protocol with the server
Validate the x509 certificate that the server sends
Telling the client that I want to work in passive mode and that i want to read a file with the command RETR.
Now, I need to open a data connection to read my file. So what I wanted to know is does the data connection is secured in the same way the control connection is ? That means do I have a certificate to validate ? Is the handshake the same ?
If this isn't clear let me know i'm not an expert at all in this area.
Thank you for your future responses
In explicit FTPS connection after you send AUTH command, the data connection is secured all the way. Depending on Active/Passive connection the data connection is opened through port 20(Active) or some random port negotiated(Passive).
You don't need any extra handshakes as far as I know.
You might or might not be able to configure if you want to reuse the same session ID created in opening the control channel. This is related for server to know that data connection has been established by the same client which opened the control channel. Some server require this, some servers allow this, some servers does not support this.
I know generally how HTTPS works, and from what I read, it seems that the generation of the encryption key could be done by a computer that isn't the one establishing the connection.
My thought is to create a proxy that isn't actually able to see the traffic going over the SSL connection. To do this, you would have a handshake in the same way as usual between the proxy and the server. However you would pass the encrypted symmetrical key to the proxy from your computer who would then transmit it on to the server. Then, when the server returned an encrypted payload, the proxy would not be able to decrypt it, but would forward it on to the users computer.
Am I missing something fundamental about how HTTPS works?
Is is possible to configure HAProxy to only allow single connection per client certificate (based on CN or other attribute)?
Each client should be able to connect even if it has already open connections, though, after successful authentication any previous connection for that certificate should be dropped immediately.
I do not think ACL's would work for this purpose.
Thanks in advance
Would like to write SSL Socket client server application on the same machine (both server and clients are running same machine). Should I use SSL certificate file (.pem file ) or not required.
Just I want encrypt the data before sending via socket.
Can we encrypt the data without .pem file?
Please advice me.
-Bala
SSL has two major steps:
check the identity of the peer, e.g. make sure you talk to the right one and not an imposter (e.g. man in the middle). This is done with certificates.
encrypt the connection, so that anybody listening on the wire cannot understand what's going on. Obviously encryption helps nothing if somebody is actively listening on the wire, because you failed in step#1. In this case you have no end-to-end encryption but an end-to-imposter and imposter-to-end encryption :)
Considering these arguments: do you still want to skip checking the certificate?
If you want it is doable, just look for anonymous authentication, e.g. ADH or aNULL with openssl.
I'm pretty far out of my depth here, so bear with me. We're using JDBC via com.intersys.jdbc.cachedriver to run stored procedures in Intersystems Cache and get the results in Java. However, there are other clients that connect directly to the Cache using a terminal over telnet. We have to lock down Cache's telnet port so that only telnet over SSL is possible. However, the Cache experts here say that locking down one port in Cache locks them all down, so the Java to Cache connection will also have to use SSL. I understand vaguely that some JDBC drivers support SSL, but I don't see that the Cache one does. What I need to know is:
What's needed to configure SSL for the JDBC connections? (We're using JBoss 4.2.3)
What about certificates? I assume those have to go somewhere.
Is it actually true that locking down the telnet connections is also going lock down JDBC in the same way?
Configuring Java Clients to Use SSL/TLS with Caché
Using SSL/TLS with Caché
Telnet vs ssh is a question about what the OS allows and only relates to Cache peripherally. It doesn't mean anything regarding items 1 or 2. So in answer to the direct question you are asking, "No".
1. What's needed to configure SSL for the JDBC connections? (We're using JBoss 4.2.3)
See http://docs.intersystems.com/cache20102/csp/docbook/DocBook.UI.Page.cls?KEY=GCAS_ssltls#GCAS_ssltls_javacli for details. This section doesn't explicitly mention JDBC, but it's true for JDBC (and the reference has been added for the documentation 2011.1 [next release] .)
2. What about certificates? I assume those have to go somewhere.
To quote the documentation: "If the client has a private key and certificate, these are stored in the client’s keystore; the keystore can also hold the client’s root CA certificate and any intermediate CA certificates. To authenticate the server, the client may need to have the root CA certificate for the server and any intermediate CA certificates, these can be stored either in the client’s truststore or along with client certificate information in the keystore. For more information on keystores and truststores, see the section “Keystores and Truststores” in the Java Secure Socket Extension (JSSE) Reference Guide."
3. Is it actually true that locking down the telnet connections is also going lock down JDBC in the same way?
Yes -- if telnet connections are to require TLS, then the superserver TLS setting needs to be Required, which means that you have to use TLS for anything that goes through the superserver (including JDBC). If you just want to allow the use of TLS for telnet connections, set the TLS value to Enabled, which allows non-TLS connections by other means (such as JDBC).