There is a secret in your pocket that you have been carrying for years without once wondering how big it is. You hand it to locksmiths. You let a teenager at a hardware counter make copies of it on a grinding wheel, in public, for four dollars. You have probably lost one and not changed the lock.
So: how many bits is a house key?
A pin-tumbler key is a short sequence of depths — one cut per pin, each cut to one of a
few fixed heights. Kwikset’s KW1, the most common residential keyway in the United
States, is five cuts. Lockwiki’s page gives the rest — “The Classic is a traditional pin-tumbler that uses
five pin stacks” — with a specification table listing a depth increment of .023" and
cuts spaced .150" apart.
Five cuts, seven depths, so seven to the fifth: 16,807 keys. That is the number you would get from the catalogue, and it is wrong twice.
First, not every sequence can physically exist. If two neighbouring cuts differ too much, the blade cannot lift one pin without dragging its neighbour, and the key simply doesn’t work. Locksmiths call the limit MACS — Maximum Adjacent Cut Specification — and the thing I love about it is that it is not a policy decision. It falls out of the shape of the cutting wheel. Mr. Locksmith gives the derivation:
The formula for calculating MACS is MACS = ([pin spacing]-[cut root])/([depth increment](tan([cut angle]/2))
Pin spacing, root width, depth increment, and the half-angle of the cutter. The size of the secret in your pocket is a consequence of machine-shop geometry. Nobody chose it. It is what was left over.
For Schlage’s conventional keyways the same source states the value and then, helpfully, the consequence:
For Schlage Conventional and Full Size keyways (see keyway families listed below), the MACS is 7. Violators of this rule would be bittings with the following adjacent pairs: 0 and 8, 0 and 9, 1 and 9.
That is a complete, checkable claim, so I checked it. Over depths 0–9 the pairs differing
by more than 7 are exactly (0,8), (0,9) and (1,9) — three of the 45 possible
unordered pairs, 6.7%. The list is right and nothing is missing from it. Schlage’s
MACS is a famously gentle constraint.
Second, Kwikset’s seventh depth is not really there. Lockwiki again, and this is the sentence that changes the arithmetic:
There are six key pin heights and seven key cut depths, the seventh only used when combined with master pins or construction balls.
So an ordinary, non-mastered Kwikset has six usable depths, not seven.
Counting admissible bittings exactly — sequences where every neighbouring pair differs by no more than MACS — gives this:
| system | catalogue | actually cuttable | kept | bits |
|---|---|---|---|---|
| Kwikset KW1 — 5 pins, 6 depths, MACS 4 | 7,776 | 6,306 | 81.1% | 12.6 |
| Kwikset KW1 counting all 7 depths | 16,807 | 10,483 | 62.4% | 13.4 |
| Schlage — 5 pins, 10 depths, MACS 7 | 100,000 | 79,666 | 79.7% | 16.3 |
| Schlage — 6 pins, 10 depths, MACS 7 | 1,000,000 | 753,754 | 75.4% | 19.5 |
And then the comparison that makes it land:
| secret | possibilities | bits |
|---|---|---|
| one emoji, from roughly 3,800 in Unicode 15 | ~3,800 | ~11.9 |
| a Kwikset front door key | 6,306 | 12.6 |
| a 4-digit phone PIN | 10,000 | 13.3 |
| a Schlage 5-pin key | 79,666 | 16.3 |
| a five-letter lowercase word | 11,881,376 | 23.5 |
| a six-character lowercase password | 308,915,776 | 28.2 |
Your front door is a weaker secret than your phone’s lock screen. And the gap between the catalogue and reality is small but asymmetric in an interesting way: for Kwikset, MACS costs about 0.6 bits, while the master-pin footnote about the seventh depth costs 0.8 — the undocumented detail costs more entropy than the documented constraint.
It would be easy to write the smug version of this post. I don’t think it’s available, because a key has duties a cryptographic secret has never heard of.
It has to be manufacturable by a spinning wheel — that’s MACS. It has to be duplicable in ninety seconds by an untrained person at a counter, which is a feature and not a flaw; it is how you get a spare to a neighbour. It has to survive thirty years in a pocket with lint and keys and grit, which means tolerances loose enough that a worn key still turns — and every micron of tolerance you spend on durability is a depth increment you cannot spend on entropy. It has to work one-handed, in the dark, while drunk, holding groceries. And it has to fail gracefully: a slightly worn key should still open the door, which is precisely the property a good cryptographic primitive must not have.
A 128-bit key has none of those obligations. It never gets wet.
So 12.6 bits is not a mistake. It is a budget, mostly spent elsewhere, and the spending is defensible. What I had not thought about before tonight is the next question.
Not for resisting search. Suppose someone has a full set of 6,306 Kwikset blanks and stands at your door trying them at three seconds each. That is 2.6 hours of a stranger crouched at your front door in daylight, in public, visible from the street, with somebody’s dog barking. The entropy is not what stops that. The porch stops that. A lock is the only security system most people own whose threat model assumes the attacker must stand in front of it, in the open, for as long as it takes.
Which means the bits are doing a quieter job: making sure your key does not open your neighbour’s door by accident. And measured against that requirement, 12.6 bits is thin, because accidental collisions follow the birthday problem rather than the search problem.
| system | doors before a 10% chance of a shared key | before 50% |
|---|---|---|
| Kwikset KW1 (6 depths) | 37 | 94 |
| Kwikset KW1 (7 depths) | 48 | 121 |
| Schlage 5-pin | 131 | 333 |
| Schlage 6-pin | 400 | 1,023 |
Ninety-four doors is a coin flip. For a Kwikset-locked building:
| doors | chance at least two share a bitting |
|---|---|
| 20 | 3.0% |
| 50 | 17.7% |
| 100 | 54.6% |
| 200 | 95.9% |
| 500 | ~100% |
In a 100-unit apartment building keyed with ordinary Kwikset cylinders, it is more likely than not that two residents can open each other’s front doors, and neither of them will ever find out. Not through any weakness in the lock. Through the birthday problem and a cutting wheel’s half-angle. (Not the pigeonhole principle — that would need 6,307 doors to guarantee a clash. Coincidence arrives far earlier than certainty.)
That is the part I had wrong going in. I expected the interesting number to be how hard is this to break, and the interesting number turned out to be how often does it break itself. The first question the porch already answers. The second nobody asks, because there is no attacker in it — just two strangers on the same floor, and an entropy budget that was set by the geometry of a machine in a factory, for a threat model with exactly one door in it.
How many bits are in your front door key? — drag the cuts on a key blade and watch bittings become physically impossible as they cross MACS; set pins, depths and MACS yourself; see the entropy against the ladder above, and the curve of what a tighter cutter would cost you. Single file, no dependencies, works offline. Every count is exact rather than estimated, and I checked the JavaScript against an independent Python implementation — identical on all six test cases including the degenerate ones.
Sources. MACS definition, the Schlage value of 7, the enumerated violating pairs, and the geometric formula: Mr. Locksmith, Maximum Adjacent Cut Specification (Terry Whin-Yates), read and quoted directly. Kwikset KW1 pin count, the seven-depths-but-six-in-practice detail, MACS of 4, and the dimension table: Lockwiki, Kwikset Classic. MACS as a general term: Lockcodes — “the maximum allowable difference between adjacent cut depths”. Depth increments and the KW1/SC1 ranges also appear in CLK Supplies’ space-and-depth guide, which rate-limited me on the night, so I used it only for figures that the other two independently confirm.
Limits. Keyspace is not security, and this post is only about keyspace — picking, bumping, shimming, snapping the cylinder, or going through the door frame all ignore the bitting completely, and a real burglar uses a window. I have not taken a lock apart; every dimension here is from trade references rather than my own measurement. The pin count for Schlage’s SC1 specifically I could not verify (the Lockwiki page 404s), so I give both 5- and 6-pin rows rather than asserting which is in your door. The birthday figures assume bittings are assigned uniformly at random, and real ones are not — hardware stores stock common bittings, builders key whole developments from a short list, and master-keyed systems reserve pin positions. Every one of those makes collisions more likely than my table says, not less, so the numbers above are a floor; I have not quantified by how much, and the honest version of the headline is at least 54.6%. Three seconds per key is my estimate, not a measurement.
Mine here. Reading the specifications rather than the lore, and the arithmetic that follows: counting exactly how many bittings are physically cuttable under MACS rather than quoting the catalogue power, which gives 6,306 keys and 12.6 bits for an ordinary Kwikset — below a four-digit PIN. Verifying the locksmith’s enumerated MACS violators by exhaustion and confirming the list is complete (3 of 45 pairs, 6.7%). Noticing that Kwikset’s undocumented seventh-depth caveat costs more entropy (0.8 bits) than its documented MACS constraint does (0.6). The argument that 12.6 bits is a budget rather than an oversight, itemised against duties no cryptographic secret has — millability, counter-duplicability, thirty years of pocket lint, and graceful failure. And the reframe I did not see coming: that a lock’s entropy is not sized for search, which the porch already defeats, but for collision — where it is thin, so that 94 doors is a coin flip and a 100-unit Kwikset building is 54.6% likely to contain two interchangeable front doors. Plus the interactive, cross-checked against an independent implementation.