I don't know of any off-the-shelf standard way of doing this. SSL could work if the proxy is a simple TCP proxy that just forwards packets and rewrites parts of the TCP header, in that case the SSL session is actually terminated by the actual LDAP server. On the other hand, if the Proxy is more intelligent then that, then the SSL session would terminate with the Proxy server and you'd lose your nonrepudiation.
It is the later case that's harder to solve and has no standards. One way to do it is to construct your system so that it uses a custom written LDAP extension, or perhaps just custom attributes in the LDAP database, that when queried returns a checksum generated from the returned statement. That would allow detection of an evil proxy. Or if you require more specific nonrepudiation, generate a value based on a preshared key of some kind rather than a checksum. Unfortunately, you'd probably have to make some changes to your LDAP server.