techbullion.com

How .onion Addresses Actually Work – and Why You Can’t Guess One

By Khizar Nisar·Published October 19, 2026·Cybersecurity

How .onion Addresses Actually Work - and Why You Can't Guess One
Fifty-six characters, no registrar and no DNS zone. Here is what is really happening when a .onion address resolves.

A .onion address looks like a typing accident: 56 characters of lower-case letters and digits in no recognisable order, with no brand in them and nothing to memorise. That is not a quirk of design. The address is not a name that points at a service – it is the service’s public key, written in a form a browser can carry. Once that is clear, most of the familiar questions stop being mysterious: why the strings are so long, why they change, why you cannot look one up, and why a search index is needed on a network that is, in principle, just websites.

The address is a key, not a name

A current (v3) address encodes 35 bytes in base32: a one-byte version, a 32-byte ed25519 public key, and a two-byte checksum. Fewer characters and the checksum fails; one character changed and it fails. So the address verifies itself. On the ordinary web you lean on a registrar and a certificate authority to establish that a domain belongs to the company you think it does. A .onion has neither. Whoever can prove possession of the private key is that address, there is nobody to appeal to, and there is no transfer, no renewal invoice and no dispute process.

Why the strings got longer

Earlier addresses were 16 characters, built from shorter key material and a hash. Tor retired that format in 2021, because the smaller key space was no longer a sensible trade-off. That history is a free diagnostic: a directory that still prints 16-character addresses has not been maintained since, and a list that mixes 16- and 56-character entries is a snapshot somebody stopped updating.

What happens when you open one

Your browser cannot resolve the address the way it resolves a domain, because there is no DNS entry to resolve. It hands the address to Tor, and four things happen:

Six relays end up between the two parties. Neither side ever learns the other’s IP address, and the rendezvous point sees two anonymous circuits that it cannot attribute to anyone. Nothing in that exchange resembles a lookup, which is why calling it “resolving” is misleading.

Why you cannot enumerate the network

There is no zone file to walk and no certificate transparency log to scrape. The key space is 256 bits wide, so guessing an address you were not given is not a matter of patience. This is the structural difference from the open web. There, a crawler finds pages by following links, and short domain names can be guessed cheaply enough to be worth trying. On Tor a service exists for you only if its address reaches you from somewhere else. Somebody has to publish it.

Which is why discovery is the hard part

In practice that means an index: a crawler that starts from published addresses and follows links between services, or a catalogue maintained by hand, or – most often – both. A search index such as OnionLand Search Engine is doing the job DNS would do if a registry existed. What matters in that map is not its size but its upkeep: whether dead entries are removed, whether mirrors are recorded, and whether the same service is listed once rather than five times under five addresses.

What an address does not tell you

“The address is valid” is a low bar. It proves control of a key, and nothing else.

The short version

A .onion address is a public key in base32: 56 characters, self-verifying, unregistrable. Opening one is a rendezvous between two circuits, not a lookup. The network cannot be enumerated, so discovery depends on published lists and on the discipline of maintaining them – which is why indexes exist, and why a good one is judged by how it handles dead, duplicated and mirrored entries rather than by how long its list is. And you need Tor Project‘s browser to reach any of it: no other software resolves these addresses.