Cold Boot Attack

A cold boot attack recovers the contents of a computer's volatile memory — RAM — shortly after power is removed, exploiting a phenomenon called data remanence. Contrary to the once-common assumption that DRAM loses its contents instantly when powered off, memory cells actually retain their charge for seconds to minutes, long enough for an attacker with physical access to read them. The attack was demonstrated in the landmark 2008 "Lest We Remember" research from Princeton, which showed full-disk-encryption keys being recovered from supposedly "off" machines — and it remains the canonical example of why physical access changes every security assumption.

How keys leak from RAM

While a system runs full-disk encryption, the decryption key must sit unencrypted somewhere in RAM — otherwise the machine could not read its own disk. A cold boot attacker cuts power, then either reboots into a tiny custom kernel that dumps the surviving memory, or physically pulls the DRAM modules and reads them in a machine they control. Because the key is plaintext in memory, recovering it bypasses the disk encryption entirely — no password guessing required. Cooling the chips dramatically slows the charge decay: spraying inverted canned air or another cooling agent extends the readable window from seconds to minutes, long enough to move modules between machines. The same logic applies to anything else resident in RAM at the moment of attack: cached passwords, session tokens, and — of direct interest to Bitcoiners — decrypted wallet keys in a running wallet's memory.

Mitigations

Defenses focus on keeping keys out of vulnerable RAM, or shrinking the window in which they exist there. Some systems hold keys only in CPU registers or dedicated hardware key stores rather than main memory; others scrub keys from RAM on shutdown, sleep, or lid-close and refuse to resume without re-authentication. Binding disk keys to a Trusted Platform Module that releases them only in a verified boot state raises the bar further, and newer platforms can encrypt memory contents themselves. Behavioural mitigations are free: power machines fully down rather than sleeping them when they leave your control, since a powered-on or suspended device with an unlocked volume is far more exposed than a cold one — after a few minutes genuinely off, there is nothing left to read.

What it means for self-custody

For a Bitcoiner, the threat model translates directly. A hot wallet on a laptop means key material transits RAM every time you sign; a seized or stolen machine that was sleeping is a live target. This is one reason cold storage keeps keys off general-purpose machines entirely, and why a hardware wallet is the right signing surface: the private key lives inside a device built to never export it, and your computer's RAM only ever sees signed transactions, not secrets. A seed phrase typed into a general-purpose computer, by contrast, has passed through exactly the memory this attack reads.

Perspective

Cold boot is a physical-access attack with real preconditions — proximity, timing, preparation — not something that happens over the network. Its enduring lesson is architectural: assume any secret that touches a general-purpose machine's memory can be captured by whoever controls the hardware, and design your custody so the secrets that matter never make that trip. That assumption costs little and closes an entire class of attack, which is exactly the kind of trade a sovereign setup should take.

The same reasoning should shape how you handle machines at borders, in repairs, or in shared spaces: a device out of your sight is a device whose memory, storage, and firmware you can no longer vouch for. Power down fully before it leaves your hands, and keep anything that signs transactions on dedicated hardware that never held the secret in general-purpose RAM to begin with. Physical custody of the machine is part of custody of the keys.