Privacy is a meta-property of the connection; "sharing" information obtained from a given message occurs outside of the original connection, so it is not strictly a property of a connection. Here's how I'd define privacy in terms of the model:
- Privacy
- Privacy is only an issue when the sender includes information in a message about the sender or some other entity that could be used to cause that entity some form of harm. "Harm" here runs the gamut from "minor annoyance" (another spammer gets your email address) through "life-threatening" (the location of a Witness Protection Program protectee is leaked), varying based on the nature of the information and who obtains it. The right to privacy is the right to control who has access to that information.
I've deliberately left the wide range of possible harm in the definition because I believe that matches how we use the term. Of course as always the exact nature of the possible harm plays into whether a given action is ethical.
In addition to data strictly contained in the message, people are also concerned about the collection and distribution of metadata about the messages, such as patterns in web page requests or what kind of television shows they tend to watch. In a way, this is information that is still contained in the messages, as it can not exist without any messages sent at all. So while it may not be directly contained in any one individual message, there is nothing special about metadata that merits special definition or handling.
Based on the clarity afforded by this definition, we can knock down another common argument against the need for privacy: "If I'm doing nothing wrong (usually wrong is used synonymously with "doing nothing illegal"), then I don't need privacy." There are two basic problems with this argument: One, "privacy" encompasses far more then just "hiding illegality"; certainly information about the commission of illegal, immoral, or socially unacceptable acts fits into the definition above quite handily, in that extreme harm can come to the entity if the information is shared with the wrong people, but that is hardly the only information that fits the definition. It is trivial to come up with instances where a person is doing nothing wrong at all, yet may still wish to prevent some other entity from obtaining information about them. For instance, there's the Witness Protection Program I used parenthetically above; the witness has not (necessarily) done anything wrong. Or consider someone being stalked who wishes to prevent a stalker from obtaining their address or other vital information. (And it's not just celebrities who get stalked, us Normals have to deal with it as well. Most of us probably know someone who has been stalked (to varying degrees) at some point.) Obviously these are extreme examples used for rhetorical purposes, but lesser examples are easily thought of, too.
The second problem is the hidden assumption that the purpose of privacy is inevitably to commit the "sin of omission", to hide something that you should be punished for. I would say this is incorrect. Let us explore the question "What is the ethical reason that privacy is desirable?"