Skip to content

Hide who, when, how

The metadata-private messenger

Encryption hides what you say. Metadata - who, when and how often - is what surveillance actually runs on. PrivaMesh hides that too.

One-time addresses

A fresh one-time address per message hides the social graph.

Cover traffic

Decoy messages hide when you actually send.

Fixed-size padding

Uniform message size hides how much you say.

Two people can exchange perfectly encrypted messages and still be fully exposed. If an observer knows that A messaged B at 2:14am, then again after B replied, they’ve learned the relationship, the rhythm, and the timing - without reading a single word. That is metadata, and most messengers leak it freely. PrivaMesh is engineered so it doesn’t. This is what makes it a genuine messenger that doesn’t collect metadata: there is no collector, and the delivery trail is deliberately unlinkable.

One-time addresses - hide who

Every message is delivered to a fresh one-time address that only the intended recipient can recognize. Two messages to the same person go to two unrelated-looking addresses. An observer sees a scatter of one-off addresses with no way to cluster them into a conversation or a social graph. The who talks to whom simply isn’t written down anywhere.

Cover traffic - hide when

Timing is its own leak. If your real messages are the only activity you ever produce, their timing reveals your pattern. PrivaMesh mixes in cover traffic - decoy messages indistinguishable from real ones - so an observer cannot tell a genuine send from noise. Frequency analysis and timing correlation lose their signal.

Fixed-size padding - hide how much

Message length leaks meaning too - a one-word “yes” and a long confession look very different if their sizes differ. PrivaMesh pads every message to a fixed size before sealing it, so a short reply and a long paragraph are identical on the wire. A separate fee payer covers any network cost, so the account paying is never the account sending.

Why this needs no server to work

None of this relies on a trusted party promising not to look. It is built into how messages are addressed and delivered. Because there is no PrivaMesh server, there is also no server-side log of IP addresses, timestamps, or contact lists - the metadata most apps leak first.

The honest trade-off

Metadata protection is strong but not magic. Your anonymity still depends on good operational hygiene - for example, the network you connect from. Good practice matters. We’d rather tell you where the edges are than pretend they don’t exist.

PrivaMesh private social graph hiding who talks to whom, no phone numbers, on iPhone

Frequently asked questions

Why does metadata matter if my messages are encrypted?

Encryption hides what you say, not who you say it to. If an observer knows that A messaged B at 2:14am and again after B replied, they have learned the relationship, the rhythm and the timing without reading a word. That is what surveillance actually runs on.

How does PrivaMesh hide who I am talking to?

Every message goes to a fresh one-time address that only the intended recipient can recognize. Two messages to the same person land on two unrelated-looking addresses, so an observer sees a scatter of one-off addresses with no way to cluster them into a conversation or a social graph.

Can someone work out when I am messaging?

Cover traffic mixes decoy messages, indistinguishable from real ones, into your activity. Frequency analysis and timing correlation lose their signal because a genuine send cannot be separated from noise.

Does message length leak anything?

It would, so PrivaMesh pads every message to a fixed size before sealing it. A one-word reply and a long paragraph are the same size on the wire.

Primary sources

Keep reading