We have a client whose code was written in Java 1.7. Java 1.7 by default refuses to connect via HTTPS to servers who return an SNI unrecognized_name warning. It's possible to turn off this behavior, but (of course) our client can't do that. Most other clients just ignore the warning.
We have a valid wildcard certificate for our domain, let's call it *.widgets.com. Anything in the domain widgets.com resolves to our HAProxy load balancer. We've installed that cert onto the load balancer, and we specify it in the front-end that listens on port 443. The cert is current and checks out fine when we test it from Qualys... except for that SNI warning.
Our client makes a call to a specific subdomain, say foo.widgets.com. The service is working fine, serving up content to anyone who calls it. Except for our client, of course, who won't connect to us after we return the SNI warning.
I've found lots of articles about how to solve this problem on Apache, but those don't help me with HAProxy. On HAProxy, I see that I can specify more than one cert, and I am told that HAProxy will "choose the right one". Do I need to get a separate, non-wildcard cert for foo.widgets.com? I don't want to buy another cert only to find out that that was not the solution.
Turns out the problem had little to do with HAProxy. Apparently we had an intrusion detection system in place that would terminate TLS prior to relaying down to HAProxy.
There is probably a way to make the IDS behave properly, presenting the correct certificate to the client. But we don't really need IDS on our non-prod environments, anyway. So we left it switched off, and the problem went away.
So if you're having a similar issue, after making sure that your certificate is good for the request you're testing, my advice would be to check whether you have any security software that could intercept traffic before it reaches your LB.
Related
I am using the Let's Encrypt IIS client from https://github.com/Lone-Coder/letsencrypt-win-simple to generate a certificate for a server. Since the certificate is only valid for three months, I want it to auto-renew.
But the server for which I need that auto-renewing certificate is only bound to https:||mysubdomain.mydomain.com:443 and smtp:||mysubdomain.mydomain.com:25.
Both http:||mysubdomain.mydomain.com:80 and ftp:||mysubdomain.mydomain.com:21 point to a different server.
As you may have guessed, the error that is now thrown during the process is "The ACME server was probably unable to reach http:||mysubdomain.mydomain.com:80/.well-known/acme-challenge/abcdefgh...xyz".
It is completely clear to me why, but I can't fix it, because http:||mysubdomain.mydomain.com has to point to the other server. If the ACME server would try https:||mysubdomain.mydomain.com:443/.well-known/acme-challenge/abcdefgh...xyz, but ignore any certificate issue, he would successfully find the challenge.
Is there anything I can do, any feature I have overlooked, that would help me to get automated renewal working?
There are multiple options:
http-01
Redirect http://example.com/.well-known/acme-challenge/* to https://example.com/.well-known/acme-challenge/*, Boulder will happily follow any such redirect and ignore the provided certificate. That's the most simple way if you have access to the other server and can configure that redirect. It's a permanent redirect that you don't have to adjust, it'll be just fine every three months.
The option to use HTTPS directly has been removed due to security issues with some popular server software that uses the first host defined if some other virtual host doesn't define any HTTP host, which might lead to wrong issuances in multi-user environments aka shared hosting.
tls-sni-01
If you want to use just port 443, you can use another challenge type called tls-sni-01. But I think there's no client for Windows available yet that supports that challenge type.
dns-01
If you have control over the DNS via a simple API, you could also use the DNS challenge, it's completely independent of the port you can use.
I have two projects running on Wildfly-8 and I have two SSL certificates for each of them and one IP.
I figured out that I should have one IP for one SSL certificate.
But I needed to use these two SSL for one IP. I couldn't find a way to do it with Wildfly but there was a way to do it with Apache Server. So,I installed Apache Server up to Wildfly.
I listen https port(443) on Apache and redirect it to Wildfly's http port(I used 8080). It works without any problem.
What I wonder is;
1. Is Apache decrypt request and redirect it to Wildfly?
2. Is it correct way to do it or I have done it by chance?
3. Does this method create a security hole?
I googled some, but I could not find satisfied answers.
Thanks for replies.
For this answer, I'm supposing that by "redirecting" you mean "proxying": Apache receives the request, proxies it to Wildfly, receives an answer from Wildfly, sends the answer to the client.
If you mean something else, then the simple answer is: it is wrong[1].
Is Apache decrypt request and redirect it to Wildfly?
Yes. Apache will receive and send secure data to/from the client. Its communication with Wildfly will be plaintext.
Is it correct way to do it or I have done it by chance?
That's how it's usually done, yes. In other words: a load balancer and/or a proxy in front of Wildfly (Apache in your case). Wildfly itself is not reached directly by the public internet.
Does this method create a security hole?
It does, just like everything else is a security "compromise". In this case, you are trusting your internal network, in the name of a more practical/manageable architecture. If you do not trust your internal network, you should look for another solution. In the general case, the price to pay seems fair to me, as you'll "only" be open to a man-in-the-middle between your Apache and your Wildfly. So, if you trust your internal network, you should trust that there won't be any MITM there.
Edit
[1] - As everything else in life, there's no absolute truth. Basically, there are 3 techniques that can be used in a scenario like this: pass through, edge and re-encryption.
Pass through is a "dumb" pipe, where nothing about TLS is known by the proxy. Wildfly would then handle the secure communication with the client. I'm not sure Apache would do this, but this can be done with haproxy in TCP mode;
Edge (or offloading) is the situation I described above: Client talks TLS with Apache, Apache talks plaintext with Wildfly;
Re-encryption, which is like Edge, but the communication between Apache and Wildfly is also TLS, using a different certificate.
I have a doubt that I had for years and now I decided to try to understand it. I know that when a user hits a website with SSL all headers are encrypted, even the HOST header.
So, in order to enable SSL in a server, you need to have a single IP to every certificate you have cause Apache, for example, wont know which VHOST it should redirect the user if the HOST header is encrypted.
My question is: how does Cloudflare knows which domain the user is using to access its CDN if it does not know the HOST before the decrypt happens?
Server Name Indication (SNI) allows TLS clients to specify the host they are attempting to connect to give the server a chance to serve the right certificate. It is supported in most browsers.
CloudFlare's page on their free SSL offering indicates they use SNI.
Now, CloudFlare has multiple offerings. Their paid plans don't actually rely on SNI (that's why they support all browsers). Only the free plans do.
For the paid plans, CloudFlare presumably uses dedicated IPs, though even in that case they can still pool multiple domains under a single certificate (using Subject Alternative Names).
I've created WCF Service and I share it via ssl. I have little knowledge about security, but I'm curious why can I see whole communication as a plain text in httpAnalyzer, even though POSTs are sending via https?
When my client application invokes wcf service, then I can see it in sniffer - passwords etc.
Does it mean that SSL works only on the lower layer - while transporting data? So every evil application can sniff communication on client's side and an encryption only secures us against man-in-the-middle?
SSL works indeed on a "lower layer" than HTTP. According to the OSI Model, SSL works on the Session Layer, while HTTP is on the Application Layer.
Most of these clientside HTTP Analyzers work from within the browser, analyzing the HTTP traffic on the application layer, before it is processed by the SSL logic. So it is completely normal to see the plain HTTP request.
Concerning security, an evil application installed within the browser can indeed read upon the traffic. But once it is processed by the SSL layer, it becomes way harder for an evil application to read the traffic.
SSL works by firstly authenticating the server to you as a client. (Do I talk to the one I really want to talk to). As you can't know all of the servers and their certificates before hand, you use some well known root certificates, which are pre-installed on your OS. These are used to check if some server is perhaps known by an already well known service. (I don't know you, but some really important server tells me that you indeed are who you say you are).
This authentication step works independent from the encryption of the traffic. No program can decrypt an arbitrary SSL stream by "installing a root certificate". (As said these root certificates are already on your machine from the first moment you install an OS on it =)
But if a evil programs is able to let you believe that you are talking to a legitimate server, using a forged root certificate for example, instead of actually talking to malware, it is able to see what the contents of the SSL traffic is. But then again, you are talking to the evil program itself, not the server you were intended to talk to. This is however not the case with HTTP Analyzer
This is in short terms how SSL works and hopefully answers your question.
Most likely HTTP analyzer install it's own root certificate, and intercepts SSL traffic, working as man-in-the-middle.
Here's a possible scenario.
Let's say I have a website "https://www.mywebsite.com" and there is a valid SSL certificate purchased for this domain.
I want to "mimick" this website on my LOCAL machine for a testing purpose.
So let's say I set up a locally-configured "https://www.mywebsite.com" (which is in essence https://localhost/mywebsite or something similar).
Would I be able to re-use the SSL certificate on my local testing website?
You can re-use your SSL certificate if you configure your DNS so that your test machine is the same domain name as server, which is probably a bad idea.
You can also re-use it on your test machine if you don't mind clicking the box "accept this whacked out ssl cert"... So I suppose that the answer is technically yes, although I wouldn't personally do it.
It depends what you are trying to test and why you need a certificate for testing.
If you use the certificate, it will correctly encrypt connections using SSL, but any client will get a certificate mis-match error. If you use a self-signed certificate instead, most clients will give you a warning about that, so it might be just as annoying or not.
If you are testing, for instance, a deployment script to make sure everything gets installed in the right place, it will work. If you are testing to make sure your code correctly redirects a non-secure connection to a secure one, it will work.
If you want to test the your website for functionality, usability, bugs, etc. then your testers will likely complain about the certificate warnings or errors, and you're probably better off doing something else.
I am not sure since the SSL certificate is bound by the domain name that was registered with the certificate. But you may be able to dupe the certificate by editing your hosts file to change localhost 127.0.0.1 to be mysite.com 127.0.0.1, ...in theory at least...if not this is a question for serverfault.com.
Hope this helps,
Best regards,
Tom.
You can't use it since the SSL cert is tied to the domain www.mywebsite.com unless you do a bit of trickery.
You can put an entry in your hosts file saying that domain is at 127.0.0.1, but that's not ideal as you could no longer reach the website.
If you just need a valid cert to test with, then a better alternative is to self-sign using the IIS Resource Kit.
I'm no expert on DNS, but this would introduce a pretty major vulnerability.
Basically if this was allowed, DNS poisoning could be used defeat the whole purpose of third party trust.
Think about it:
I infect your computer so that when you go to www.amazon.com it resolves www.amazon.com to a different domain. That domain uses amazon's ssl cert to fool you into thinking it's legit, so you send me your credit card information.
So, the answer to your question is, no you can't do this. You will still get errors, My guess is that somewhere on the verfication chain, it compares the domain that initiated the request with what its internal dns resolves the domain, to verify there is a match.
As others have said, you can test SSL with a Self Signed Cert, you just have to instruct your testers to import the cert, or go through the trouble of building your own trusted CA, and have testers add that CA as a trusted CA.
There is no point in stealing another sites SSL Cert.
Of course you could use the vulnerability in MD5 to create your own valid SSL cert.
http://www.digicert.com/news/2009-01-05-md5-ssl.htm