“The Byzantine Generals Problem,” by Leslie Lamport, Robert Shostak, and Marshall Pease, appeared in ACM Transactions on Programming Languages and Systems in 1982. It introduced the now-standard allegory of a group of generals, some of whom may be traitors, who must agree on a common plan of action using messengers, even though the traitors will lie and send conflicting orders to confuse the loyal ones.
The paper recasts an earlier, more formal result on reaching agreement in the presence of faults into a memorable story, and the word Byzantine became the accepted name for arbitrary, possibly malicious failures. The central theorem is that loyal generals can reach agreement only if more than two-thirds of the participants are loyal: with oral (unsigned) messages, agreement is possible if and only if there are more than three times as many generals as traitors.
The authors also show that the situation changes with cryptographically signed messages. Unforgeable signatures prevent a traitor from misrepresenting what another general said, which makes agreement achievable with a smaller fraction of honest participants, anticipating the role digital signatures play in later Byzantine protocols.
The paper is one of the foundational works of fault-tolerant distributed computing. Its three-to-one bound and its vocabulary underpin everything from spacecraft control systems to modern blockchain consensus, and it remains a primary reference for anyone reasoning about agreement under adversarial conditions.