History and attacks

2010: the value overflow

In August 2010 someone created a transaction whose outputs summed to more than the maximum a signed integer could hold. The check that outputs must not exceed inputs passed, because the sum had wrapped around. A block containing it created roughly 184 billion BTC.

Within hours a patched client was released, nodes upgraded, and the honest chain overtook the bad one, the bad block was reorganised away. What this proved: the rules live in software written by people, and the recovery mechanism is people running the corrected software, fast.[27]

2013: two versions, two chains

In March 2013 a large block was accepted by newer nodes and rejected by older ones, because of a difference in the underlying database, not in the stated rules. The network split for a few hours.

The resolution was coordinated: miners were asked to return to the older behaviour temporarily so the chain reunified, and the upgrade was rescheduled. BIP 50 documents the whole episode, including what was learned about undocumented implementation limits being part of consensus in practice.[29]

2018: the bug that was not used

CVE-2018-17144 allowed a transaction spending the same input twice to crash nodes, and in some versions to inflate supply. It was reported privately, patched, and announced with an upgrade urgency notice. As far as the record shows, nobody exploited it.

This is the mundane, important part of Bitcoin's security: an open codebase, people who read it, disclosure that goes to maintainers first, and operators who upgrade. Not magic, maintenance.[28][2]

Attacks people worry about, and what they can actually do

51% of hashrate. A majority miner can reorder or censor recent transactions and double-spend its own coins. It cannot steal coins from an address, change the 21 million limit, or create invalid blocks that nodes will accept. Those limits are enforced by validation, not by hashrate.

Eclipse and network attacks. Isolating a node from honest peers can feed it a false view of the chain. Running your own node with multiple connections, and waiting for confirmations, is the practical answer.

Social and governance attacks. Historically the harder fights were about rule changes: the 2015–2017 block size dispute ended with users choosing which software to run, and the chain that most economic users validated remained Bitcoin. That is the real governance mechanism, voluntary validation, not a vote.

Quantum computing. Today's machines are far from breaking secp256k1 or SHA-256. It is a genuine long-term research question with proposed migration paths, not an imminent event. Anyone selling urgency about it is selling something.[1][2][4]