The formal definitions
NIST's Blockchain Technology Overview (NISTIR 8202, October 2018) is the clearest US government reference. In a permissionless network anyone can publish a new block without permission from an authority. In a permissioned network, block publishing is limited to users authorized by some authority, centralized or decentralized.[1]
Hyperledger Fabric's documentation frames the same line from the enterprise side: in a permissionless chain virtually anyone can participate and every participant is anonymous, while permissioned ones run among known, identified and often vetted participants under a governance model.[2]
NIST adds a nuance worth keeping: permissioned systems may still let anyone read the chain, and some let anyone submit transactions, while only the publishing step is restricted. Permissioned therefore describes a spectrum, not one design.[1]
Why permissionless chains need proof of work or stake
If anyone can join, anyone can also create thousands of fake identities. Bitcoin solves this by tying votes to computing power: the whitepaper's rule is one CPU, one vote, with the longest chain carrying the most proof of work.[3]
Ethereum ties votes to capital instead. A home validator must deposit at least 32 ETH, and ethereum.org warns that malicious offenses are punished by slashing part of that stake. The economic cost replaces the identity check a permissioned system would perform.[4]
This cost is also why public chains are slower per transaction than a closed network: consensus has to tolerate thousands of unknown and possibly hostile participants.
Proof of work and proof of stake price participation differently. Mining costs electricity and hardware on an ongoing basis; staking locks capital that is returned when a validator exits, minus any penalties. Both make it expensive to fake a majority, which is the only property that matters for security.
Neither system checks who you are. A permissionless chain does not know or care whether a validator is a person, a company or a pool of thousands of stakers, as long as the deposit or the work is real.[4]
How permissioned networks reach agreement
With known members there is no need for mining. Fabric uses pluggable consensus, supporting crash fault tolerant ordering and Byzantine fault tolerant ordering depending on the deployment, and does not require a native cryptocurrency.[2]
Identity is handled through a pluggable membership service provider, so every transaction traces back to a named participant. Fabric's docs note that when something goes wrong, the guilty party can be identified and the incident handled under the governance model. That suits regulated workflows and is incompatible with the anonymous access of a public chain.[2]
Performance follows from the smaller, trusted set. Fewer nodes need to agree and each node is known to be reachable, so ordering can be fast and cheap. The saving comes from trust, not from better cryptography.
Fabric also offers channels and private data collections, which limit certain contracts and data to named members of the network, a pattern with no direct equivalent on a fully public chain.[2]
Side-by-side comparison
The practical differences follow from who controls membership.
Read the table as a set of trade-offs rather than a ranking. Each row where permissioned looks simpler is a row where a governing body has taken on a job that a public protocol hands to economics.
| Property | Permissionless | Permissioned |
|---|---|---|
| Who can produce blocks | Anyone meeting protocol rules | Only authorized members |
| Sybil resistance | Proof of work or proof of stake | Member identity and certificates |
| Native token needed | Yes, to pay fees and secure consensus | No, in platforms such as Fabric |
| Participant identity | Pseudonymous addresses | Known organizations |
| Who can change the rules | Node operators adopting new software | The governing members |
Worked example: what joining costs
Joining Ethereum as a block producer requires no application: 32 ETH, a computer and client software. Since Pectra, one validator can hold up to 2,048 ETH, and stakers running several validators can aggregate them into one.[5]
Joining a permissioned Fabric network requires the opposite: admission by the members and an identity recognized by the network's membership service provider, and nothing to stake. The first model prices entry in capital; the second prices it in trust and paperwork.[2]
The ongoing costs differ too. An Ethereum validator earns rewards while it is online and attesting and pays penalties for downtime, so uptime has a direct price. A Fabric member bears only its own hosting cost and whatever fees the consortium agreement sets.[4]
That makes the permissioned model cheaper to operate per transaction, but it also means the network ends when its members lose interest. A public chain continues as long as anyone anywhere finds it worth running a node.
Layer 2s and the gray zone
Many Ethereum rollups are permissionless for users but depend on the nodes that order transactions, often called sequencers. Ethereum.org's scaling documentation notes these may be run by users, by a third-party operator or by a large group, and that rollups post their data to layer 1, where it is secured by Ethereum consensus.[6]
So the label depends on which right you ask about: using, reading, ordering or finalizing. A project describing itself as permissionless should say which of those four it means.
Some layer 2 teams have published plans to decentralize their sequencers over time. Until that happens, check the project's own status pages for whether users can force a transaction through layer 1 if the sequencer ignores it.
See also: What are the types of blockchain? · Glossary: layer 2
Censorship and finality in practice
The deepest difference is who can refuse your transaction. On a permissionless chain, a transaction ignored by one block producer can be included by the next, because the set of producers is open. Bitcoin's whitepaper describes nodes broadcasting transactions to all peers and accepting any valid block, not blocks from a list.[3]
On a permissioned network, the owner or consortium can admit publishing nodes and remove them, which NIST describes as a great amount of control. That is a feature in regulated settings, where a court order or an auditor's finding must be followed, and a risk for anyone who needs assets no operator can freeze.[1]
Finality also differs. Bitcoin offers probabilistic finality: the whitepaper shows that the chance of an attacker rewriting a block drops exponentially with each block added after it, which is why recipients wait for several confirmations.[3]
Ethereum's proof of stake finalizes blocks at epoch checkpoints once two-thirds of staked ETH votes for them. Reverting a finalized block would cost an attacker at least one-third of all staked ETH.[7]
None of these properties is better in the abstract. Each is a trade between openness and control, and the right one depends on who must be able to trust the ledger without trusting its operators.
The bottom line
Use a permissionless chain when the parties do not know each other and need assets that nobody can freeze; use a permissioned ledger when every participant is known and the goal is a shared audit trail. Mixing them up leads to either paying for censorship resistance you do not need, or trusting a gatekeeper you did not notice.
Educational content, not financial advice. Crypto assets are volatile; do your own research.