On 03/01/18 10:15, Yuri wrote:
03.01.2018 02:13, Amos Jeffries пишет:
On 03/01/18 02:48, Roberto Carna wrote:
Dear, I've setup a Squid transparent proxy + Squidgard on pfSEnse 2.4
in order to filter HTTP and HTTPS web content for different types of
WiFi clients on my company:
- Android (different versions)
- Notebooks Windows 7/10
- Iphone
- Etc.
In some cases, depending on the device Operating System, some apps
experiment problems, for example Facebook and some others.
The main cause of these problems is that when the same vendor is
authoring both the server software and the client "app" (or client
device OS). They can (and often do) hard-code TLS certificate checks
into their client code to detect and immediately fail in the presence
of MITM in the encryption.
Following that, SSL-Bump is still very much an ongoing project.
Selecting even a slightly older Squid version can lead to TLS features
not being supported. So when problems occur the best option is still
to upgrade to the very latest release before debugging further - today
that would be squid-4.0.22 beta.
Which is the best solution in order to setup a TRANSPARENT proxy
service in a heterogeneous scenario with diferenbt types of devices,
and running in the best mode with the minimum number of problems???
The _only_ solution is not to decrypt such traffic (the splice
action). How you determine which traffic is having such special trust
given to it is up to you. The TLS SNI is provided by the peek action
at SSL-Bump step 1.
Well, you can do it when you want. For example, take a look (for example):
https://stackoverflow.com/questions/4461360/how-to-install-trusted-ca-certificate-on-android-device
or on this:
http://wiki.cacert.org/FAQ/ImportRootCert
Or, in your case, you can differentiate users by, for example, by
network and pass wifi users to splice rule. Much approaches exists.
NP: none of which work when the application is checking the fingerprint
of the CA certificate against a hard-coded value defined by the vendor.
These are simply ways to make the SSL-Bumping process work without
"untrusted CA" warnings *if* bumping is already possible.
Or do I have to move to a scenario with a defined proxy in another
server, and automatically established in clients with DHCP ???
Explicit proxy is definitely better for HTTP than interception proxy.
That is true regardless of what else is going on. So worth doing *if*
you can.
Oh, really? You have forgotten about beatiful 252 option in DHCP, about
WCCP interception, and such other good things. In other world,
No, I am not.
DHCP, WPAD/PAC, and such are just ways to auto-configure explicit proxy
*not* interception proxy.
WCCP is just a way to tunnel packets to the proxy machine. Explicit
proxy does not require that additional tunnel complexity.
If you combine WCCP with interception it becomes "interception proxy",
not "explicit proxy".
Amos
_______________________________________________
squid-users mailing list
squid-users@xxxxxxxxxxxxxxxxxxxxx
http://lists.squid-cache.org/listinfo/squid-users
