BR Academy
Intro to Blockchain: How It Works and Why It Matters
0%
Exit course
Consensus and Security·Lesson 17 of 25·12 min

What Is Consensus?

The Concept

A blockchain has no central server, no admin, and no CEO who gets the final say on what happened. Thousands of independent computers (nodes), scattered across the globe and run by strangers who don't trust each other, somehow all agree on a single, shared version of history: which transactions happened, in what order, and which account owns what. That agreement is called consensus, and it's arguably the hardest problem a decentralized network has to solve.

Why not just take a majority vote among nodes? Because on an open, permissionless network, anyone can run as many nodes as they want for almost no cost. If "one node, one vote" decided the truth, an attacker could spin up ten thousand fake nodes on a single laptop and simply outvote everyone else — a flaw known as a Sybil attack, named after a book about multiple personality disorder. A voting system is only as strong as the cost of casting a fraudulent vote, and free votes mean free fraud.

This is the insight behind every real-world blockchain consensus mechanism: don't count nodes, count something costly. Proof of Work (covered next lesson) makes votes cost electricity and specialized hardware. Proof of Stake (the lesson after) makes votes cost locked-up capital that can be destroyed for misbehavior. Both approaches turn "how many voices agree" into "how much real-world value backs this version of history" — and real-world value is expensive to fake at scale.

The classic thought experiment for why this is hard is the Byzantine Generals Problem. Imagine several generals surrounding a city, who must all attack at the same time to win, or all retreat — any mismatch and they lose. They can only communicate by messenger, and some generals (or messengers) might be traitors sending false information. Can the loyal generals still reach agreement despite the traitors? This is exactly the problem a blockchain solves: nodes (generals) must agree on one version of events (attack or retreat) even though some participants might be faulty or malicious, and the only coordination channel (the network) isn't fully trustworthy either. A consensus mechanism is a mathematically provable strategy for the generals to coordinate despite the possibility of betrayal, so long as the traitors are a minority.

It's worth being clear about what consensus actually guarantees and what it doesn't. Consensus guarantees that honest nodes agree on one ordered history of valid transactions — it does not guarantee that history is "fair" in any moral sense, or that fees are low, or that the network is fast. A blockchain can have perfectly sound consensus and still be slow, expensive, or governed in ways you disagree with; consensus is specifically the answer to "whose version of events do we all accept," not a cure-all for every design trade-off a network makes. Keeping these separate matters for the capstone project later in this course, where you'll be asked to evaluate a real protocol's consensus mechanism on its own terms rather than conflating it with unrelated criticisms.

It's also worth noting that consensus mechanisms aren't just a blockchain curiosity — the underlying problem (how do independent, possibly unreliable parties agree on shared facts without a trusted referee) shows up anywhere multiple parties need a shared source of truth without a central authority: distributed databases, aviation systems with redundant sensors, and multi-party financial settlement all wrestle with versions of the same challenge. Blockchains are simply the first widely deployed systems to solve it in a fully open, permissionless setting, where literally anyone, anywhere, with no vetting or sign-up process, can join as a participant — which is exactly what makes the Sybil attack risk real in a way it isn't for a closed system with a known, vetted list of participants.

Worked Example

Picture ten generals around a city, one of whom is secretly a traitor working for the enemy. The honest generals agree in advance on a rule: "attack only if a majority of messages you receive say attack." Each general sends their vote to every other general by messenger. The traitor tries to sow confusion by telling some generals "attack" and others "retreat." But because every honest general compares the majority of votes received, not just one message, the traitor's single conflicting vote gets outvoted by the nine honest, consistent votes. The honest generals still reach the same correct decision despite one liar in the group — as long as liars stay a minority.

Blockchains apply the same logic at scale: instead of ten generals, there are thousands of nodes; instead of "attack or retreat," the message is "here is the next valid block." Proof of Work and Proof of Stake are both specific, battle-tested answers to "how do we make sure the honest majority's version wins, even when some participants are malicious?" — by making it expensive to be a dishonest participant with enough weight to matter.

Exercise

In one or two sentences, explain why requiring participants to "spend something real" — electricity in Proof of Work, locked-up capital in Proof of Stake — to get a vote prevents an attacker from simply creating thousands of free fake identities to win a vote. Then write a second sentence connecting this back to the Byzantine Generals Problem: what role does "cost per vote" play in keeping the traitors a minority of the weighted vote, even if they could technically be a majority of raw node count?

Recap

  • Consensus is how a decentralized network with no central authority agrees on one shared truth, despite some participants potentially being faulty or malicious.
  • Simple majority voting fails on open networks because fake identities (nodes) are nearly free to create — a Sybil attack.
  • Proof of Work and Proof of Stake both solve this the same way: by attaching a real, non-fakeable cost to each participant's influence over the vote.

Resources

← Reading a Block Explorer Proof of Work Explained →