How unbreakable is Signal encryption, really?
There’s a stubborn belief that encryption is only ever a speed bump: give a government or a big enough computer enough time, and any message can be cracked. For a weak password, that’s often true. For the encryption behind Signal, the same maths Fenzly uses, it isn’t. The honest answer is stranger and more absolute than “it would take a long time,” and it’s worth seeing the numbers.
What “cracking” would even mean
In the encryption post we described two things: a public/private key pair that agrees on a secret, and a lock that seals each message. In the Signal protocol those are, roughly, a key agreement built on a curve called X25519, and a message cipher called AES-256.
Neither has a known shortcut. After decades of the world’s best cryptographers trying to find one, the only known way in is the dumbest one: guess. Try keys until one works. So the real question isn’t “how clever is the attacker”, it’s “how many guesses would they need.” That’s where it stops being an engineering problem and becomes a physics problem.
A number too big to picture
The message key is one of 2²⁵⁶ possibilities, about 10⁷⁷, a 1 followed by 77 zeros. That’s close to the estimated number of atoms in the observable universe. Not the atoms on Earth. Every atom in every star in every galaxy we can see.
Analogies help more than digits. The Bitcoin network is the largest concentration of computing power humanity has ever assembled, millions of purpose-built machines doing nothing but guessing, around 10²¹ attempts every second. Point that entire global machine at the weaker of Signal’s two numbers, 2¹²⁸, and it would still grind for something on the order of the age of the universe, billions of years, before exhausting the possibilities.
And that’s the small number. For the full 2²⁵⁶ message keyspace, even the world’s fastest supercomputer as of 2026, LineShine, at about 2.2 quintillion operations a second, checking one key per operation would need roughly 10⁵¹ years. The universe is about 1.4×10¹⁰ years old. You’d be waiting around 10⁴¹ times the age of everything.
Faster computers don’t rescue you
The natural objection is: computers keep getting faster, so isn’t this just a matter of time? No, and this is the part that surprises people. The wall isn’t the speed of today’s chips. It’s thermodynamics.
Physics sets a hard floor, the Landauer limit, on the minimum energy needed to flip a single bit. Multiply that floor by 2²⁵⁶ and you get a staggering result: just to count from zero to 2²⁵⁶, not to compute anything, merely to tick a counter through every value, would take more energy than the Sun will radiate in its entire remaining lifetime, many times over. A faster processor doesn’t help, because the problem was never about processor speed. There isn’t enough usable energy in our corner of the universe to run the counter, let alone do the work.
That’s why cryptographers say a well-chosen 256-bit key is safe not “for now” but “against brute force, period”, until computers are built from something other than matter and run on something other than energy.
So why do we keep hearing about “hacked encryption”?
Because essentially no one attacks strong encryption head-on. They go around it.
Real breaches almost never crack the maths. They steal an unlocked phone. They trick someone into installing malware that reads messages after they’ve been decrypted on the screen. They guess a weak backup passphrase. They compel a person to unlock a device. The encryption is the strongest link in the chain, so attackers pull on the others.
That reframes what actually keeps you safe. It isn’t a bigger number; the number is already absurd. It’s the boring things: a phone that locks, software you trust, and a backup protected by a passphrase worth the name. The maths has your back. The rest is up to how the app and you handle the keys.
The one honest caveat: quantum computers
There’s exactly one asterisk, and it deserves a straight answer.
A future large-scale quantum computer would change two things unevenly. Against the message cipher, the best known quantum attack (Grover’s algorithm) only halves the effective key length, so AES-256 would still stand at a 128-bit level, which, as we just saw, is itself far out of reach. The symmetric layer is fine.
The key agreement is the exposed part. A powerful enough quantum computer running Shor’s algorithm could break curve-based key exchange like X25519 and the same would be true for the RSA and elliptic-curve cryptography that secures most of the internet today. This isn’t a Signal weakness; it’s an industry-wide horizon. No such machine exists, it may be many years away or never arrive at useful scale, and the wider Signal ecosystem has already begun layering post-quantum key exchange on top. It’s a future we’re tracking closely, not a lock that’s open today.
Strong keys are worth nothing if they’re lost. Next: why turning on backup in Fenzly isn’t optional housekeeping, it’s the difference between keeping your account and losing it for good.