We have many mechanisms to provide confidential communications so that network operators and other would-be surveillance regimes can't inspect the content of our traffic. But some of those mechanisms actually reveal more about who is speaking than cleartext communication would, especially over longer periods of time and large datasets. Information about who is speaking to whom is so valuable that large organizations devote huge amounts of resources to assembling network graphs of this "metadata," even without the content of the communications. Clearly this information is worth something; it is probably worth protecting. Why should privacy (hiding who you are) conflict with confidentiality (hiding what is being said)? This talk will look at specific instances of privacy and confidentiality conflicts, and describe patterns that create this tension. There will also be a discussion on some approaches to resolve the conflict and outline ways to improve privacy while preserving confidentiality.