> For example, if a user suspected that we were using a transparent proxy > to limit their access, they can have a different proxy setup on a remote > site and enter that address as the proxy address in their browser.
One reason I don't like transparent proxy is the "sneaky" aspect of the whole thing, plus you end up having to permit unproxied access to TCP/443 for SSL, and then dealing with legitimate web services running on arbitrary high ports, etc. In the end, it's just (IMHO) easier to go non-transparent. I always recommend requiring explicit proxy configuration via PAC/WPAD, and blocking all other paths to the Internet, and watching the logs for users/apps trying to tunnel out through the proxy. Every modern PC/MAC/Unix graphical browser can use Netscape's standard Proxy Automatic Configuration. We support approximately 25K users and have a default deny policy for all ports and protocols towards all Internet destinations, with minimal user complaints. For HTTP/HTTPS, in over six years we have had to configure exactly two exceptions for applications unable (unwilling) to use a proxy for HTTP/HTTPS requests. > Henrick is right though, Im not sure there is a good > way to do it with a transparent proxy. Ive been thinking about > using a radius server or NTLM (I think) and making everyone have > a username and password to get to the internet. But, that would > be quite a nightmare to set up. Another advantage of explicit proxy is the better/cleaner handling of proxy user authentication. > And finally, if that isn't sufficient, build a whitelist of allowed > sites and block everything else.. This is really the only truly effective solution. I tried this once when I was working in a SEC-regulated environment where controlling communications was important, and it worked fine (except for the ever-growing list of exempt "VIP" users.) Whitelisting also mitigates most worms/trojans/spyware/web bugs, and makes using Internet proxies to go around the filter all but impossible. Kevin