یک warrant canary چیست، و چرا اهمیت دارد؟

The warrant canary depends on a specific asymmetry in compelled-speech doctrine. Many surveillance statutes — 18 USC §2709 in the United States (the National Security Letter authority); analogous provisions in the UK Investigatory Powers Act 2016 (§133 non-disclosure); equivalents in other 14-Eyes jurisdictions — include non-disclosure clauses that prevent the recipient of an order from talking about it. Those clauses compel silence. They do not, in most jurisdictions, compel the recipient to actively make a false affirmative statement.

That distinction is what the canary exploits. A provider publishes "we have not received an NSL as of [date]" on a fixed cadence — weekly, monthly, quarterly. As long as the statement remains true, publication continues. If the provider receives an order, they cannot say so directly (gag order). But they can stop publishing the affirmative statement. The next scheduled canary either does not appear, or appears with the relevant clause removed, or appears with different language. That absence — that change — is information that the gag order cannot suppress.

The Cryptoservers canary, published at /canary/, follows this pattern. Every Monday we re-sign a statement covering the week ending the prior Sunday. The signed statement enumerates: NSLs received (target: zero), gag orders received (target: zero), bulk-retention orders received (target: zero), per-jurisdiction order counts (Saint Kitts and Nevis, Iceland, Netherlands, Romania, Switzerland), DMCA notices received versus actioned, and abuse complaints received versus actioned. Five of the six numbers are expected to be zero in a normal week; the abuse and DMCA numbers are non-zero by design (we receive abuse complaints, and we publish how many).

The cadence matters. A canary published once a year carries less signal than one published weekly — the lag between coercion and the customer noticing is the freshness window. Weekly canaries reduce that window to seven days; daily canaries (rare, because they require key-handling at scale) reduce it to twenty-four hours. Quarterly canaries leave a three-month window in which a provider could be compromised before any external observer notices.

۱. نظریه حقوقی به منفذهای قانونی وابسته است. The canary exploits a doctrine that gag orders compel silence rather than affirmative falsehood. That doctrine has not been definitively tested for every surveillance statute in every jurisdiction. The Stanford Law Review and other legal scholars have argued both sides. A determined prosecutor could plausibly argue that compelling continued affirmative publication of "we have not received" is itself part of a non-disclosure obligation, and that ceasing publication is a breach of the gag. The legal theory has held up so far in published US case law, but it has not been litigated to definitive resolution. Other jurisdictions are even less tested.

۲. canary های امضانشده بی‌فایده‌اند. A canary that is not signed with a verifiable PGP key — or that uses a key whose fingerprint cannot be cross-referenced against multiple independent sources — is theatre. An adversary who controls the provider's website can substitute a forged "everything is fine" canary and there is no way to detect the substitution. The PGP signature is what makes the canary load-bearing. Unsigned text is not a canary, no matter what it says. Verify before you trust.

۳. canary های منقضی مبهم هستند. When a canary stops being renewed, the interpretation depends on what else is happening. A lapsed canary plus organisational silence plus a key change is strong signal. A lapsed canary alone is weak — it could mean the engineer is on vacation. A canary that never started is no signal at all. The burden of disambiguation falls on the customer; the provider can only honestly publish what they can honestly publish.

Treat canaries as one cultural-integrity signal among several, not as a security guarantee. A canary tells you that a provider thinks coercion is a real enough risk to invest engineering effort in pre-publishing the contradiction signal. That is meaningful. It is not the same thing as cryptographic proof.