What the server sees anyway

End-to-end encryption protects the content. It protects almost nothing else, and this distinction is all too often overlooked.

Metadata

Even without reading a single message, the server can observe:

What the server observes	In what form
Who is writing to whom	Account identifiers
When	The timestamp of each message
How much	The size and frequency of exchanges
From where	The IP address, device type

This data is not encrypted, as the server needs it to route the messages.

Why this is far from trivial

It is easy to believe that, without the content, this information is harmless. This is not true, and an example illustrates this better than a lengthy argument.

Suppose the server observes: a message sent to an oncology clinic, then to a relative, then to an employer, late in the evening, several times a week. It hasn’t read any of the messages. Yet it knows a great deal.

Metadata also lends itself perfectly to automated analysis: it is structured and consistent, whereas the text itself would require interpretation.

What services are trying to do

Some protocols minimise these traces. Sealed sender encrypts the sender’s identity so that the server only knows the recipient. Other approaches add delays or dummy traffic.

But the protection remains partial: the server still needs to know where to deliver the message.

Key points to bear in mind for your own choices

End-to-end encryption is excellent protection for the content. If your priority is simply the fact of being in contact with someone, it is not enough.

A statement such as ‘we cannot read your messages’ is probably true. It does not mean ‘we know nothing about you’.