Can a file’s encryption change by itself?
Could the new mathematics make a file whose encryption changes over time with nothing running? Under a strict definition the requirements appear incompatible, and every way round that needs an assumption outside the mathematics.
Take an ordinary digital file. Can its actual encrypted bytes, or what it effectively decrypts to, change over time using only mathematics, with no running process, no external clock, no energy source and no physical change in the storage medium?
For defence that would be attractive: a file that expires or mutates on its own may be useless to an attacker after a while. So I wrote the question down with strict requirements and tried to answer it before getting excited.
The baseline obstacle
If every input is fixed (the file, the key, the parameters, the environment), then a deterministic decoder gives one fixed output. Swapping in a family of decoders indexed by time, or a time-varying key, changes the inputs: that's a clock or a state, which are exactly the things the strict version forbids.
For the file's physical bits to change, something physical has to change them, and mathematics alone doesn't supply a physical state transition. Randomised decoding needs entropy and computation. So under the strict definition the requirements appear incompatible. My note calls this a baseline obstacle, not a proof that every physical implementation is impossible.
The state-copying adversary
I then screened many of the manuscripts for anything that might get around the obstacle. The same reply appears in entry after entry, in slightly different words: a state-copying adversary defeats it. If an attacker can copy the file and its initial state, then however the mathematics evolves, they keep a copy of the earlier state. Decay is not erasure.
Several theorems about static structures, regularity, rationality or equidistribution, don't change this, because a static theorem has nothing evolving in it.
Where a real idea would have to look
The negative result says what a working idea would need: an assumption outside the mathematics, such as an external source of trust or a physical process. Time-lock puzzles and self-updatable encryption are the existing families that do exactly that, and I classify them as such.
Sources and notes (8)
- Le Gallo et al., Collective Structural Relaxation in Phase-Change Memory Devices (2018) — physical drift as a possible source of change
- Gao, Al-Sarawi and Abbott, Physical unclonable functions (2020) — device fingerprints and their limits
- Rivest, Shamir and Wagner, time-lock puzzles — timed release, which needs an active operation
- Time-lock puzzles in the random oracle model (CRYPTO 2011) — timed release, which needs an active operation
- openai/math — the manuscript collection screened for loopholes
- Passive encryption feasibility — the notes this was written from
- Finite passive state concept — the notes this was written from
- Passive media literature scan — the notes this was written from
Cryptography researchAll 10 parts →
