Private interactive continuity

The research agenda

Content secrecy is only one layer. Private introduction, offline delivery, presence, mobile wake-up, abuse control, recovery and live media can each recreate an identifier or reveal a relationship. The strongest proposed profile needs a composed account of all those surfaces.

A fixed schedule can let real traffic replace cover instead of creating a unique event. Private retrieval can hide which record is read. Separated operators can reduce the view held by one party. Each mechanism has a workload, failure behavior and trust premise; none supplies a complete system-level bound by itself.

The private-board prototype

PIC-10b studies a two-server private-board construction using distributed point functions. Its declared non-collusion assumption and its concrete retrieval primitive are part of the claim. The project records a desktop wire/accounting ledger of 285,113,280 bytes per declared month under its frozen workload, about 285.1 MB or 271.9 MiB. This is project-reported research evidence, not a mobile deployment measurement.

A board parameter around 2^19 addresses describes record positions. It does not establish 524,288 users, independent anonymity participants or one record per person. Capacity, active population and effective anonymity set need separate definitions.

A simpler baseline

For comparison, consider every client downloading every raw cell in a cohort. Under an illustrative workload of 1,000 users, 30 messages per user per day, 30 days and 256 bytes per message:

B = users × messages/user/day × days × bytes/message
  = 1,000 × 30 × 30 × 256
  = 230,400,000 bytes per client per month
  = 230.4 MB ≈ 219.7 MiB

At 100 messages per user per day the same raw-cell calculation is 768 MB. These are calculations, not benchmark results. They exclude upload traffic, framing, cover, manifests, authentication, retries and deployment overhead. The 230.4 MB number cannot be ranked directly against PIC-10b’s more inclusive ledger.

Full download avoids a two-server private-read non-collusion premise for the downloaded cells, but does not automatically hide writers, membership or timing, nor prevent equivocation and rollback. It is a baseline to account for, not a complete alternative system or a claim that a cohort is a fixed factor smaller than a board.

Open proof work

The record contains extensive frozen experiments and candidate classical arguments. Its adaptive quantum transition and concrete fixed-SHAKE256 justification remain open. One scoped theorem-interface calculation reports a bound around 2^-13.5 in its declared monthly pressure case, far from the intended 128-bit objective.

A weak upper bound is not an attack achieving that advantage. It means the attempted reduction does not establish the target security. Successful algebraic tests or a byte ledger do not fill that proof gap. Integration remains gated on the primitive and system composition, with independent review still required.

The phone is part of the system

Event-timed notifications expose activity to the push provider. Account tokens can link otherwise separate mailboxes. Suspension and background execution restrict constant-rate schedules. Recovery and multi-device features can reintroduce stable identifiers. Calls expose sustained, low-latency flows and have substantial bandwidth and battery constraints.

These are reasons to state a narrower observer model and measure a residual, rather than promise invisibility. The released beta’s current leaks are in the observer ledger; none of the stronger research goals should be read back into that product.