This is quite a general question:

I am trying to detect whether people accessing my network are using a proxy server. This is *not* to ensure that web pages are not cached!

The only way that I can think of doing this is to intercept at packet level and examine the source port for high-numbers. But even this is no guarantee. Humph.

Does anybody have any suggestions. The solution must be automated, ie. I can't rely on users filling out a form correctly!

No suggestions. I cannot think of any 100 percent reliable way to check this. Proxies, from the client perspective, look the same as clients. As you suggest, high port numbers are not an indication of a proxy server. Clients use high port numbers as well.

You might be able to guess proxies from large companies that direct the vast majority of traffic thru proxy servers by a study of the IP addresses of the connections. This would be a good guess for large organizations. You can also tell clients that are more-than-likely not proxied because of the hostname, i.e. the DHCP names of clients with Roadrunner, @Home, etc. This would be intelligent guessing (at best) and incomplete.