#866 1992 · IBM Almaden Research Center (Cynthia Dwork & Moni Naor) · Cryptography / computer security
Two IBM cryptographers tried to kill junk mail by charging each email in CPU work
the problem
Bulk email cost nothing to send, and no filter could keep up with senders forging addresses for free
background
By the early 1990s, email was becoming the internet's killer application — and its cheapest abuse channel. Sending a message cost the sender essentially nothing, forging a return address cost even less, and the recipients paid the difference in attention and storage. Dwork and Naor's 1992 paper named the coming flood plainly: junk mail. Every defense then imagined was a detection problem — recognize the junk, list the abusers, filter the content — and every one of them arms-races against senders who change their patterns faster than filters can learn them.
Cynthia Dwork and Moni Naor worked at IBM's Almaden Research Center in California, at a moment when cryptography was beginning to ask what economics could do for security. Their paper, 'Pricing via Processing or Combatting Junk Mail', was presented at CRYPTO '92 — and proposed to stop asking who the sender is, because identity on the network could not be trusted anyway.
what everyone would do
Filter harder: pattern-match the junk, blacklist the offenders, ask recipients to report it. Every variant treats abuse as a detection race, and every one loses the same way — the attacker changes the pattern, rotates the address, and retries at zero cost, while the defender must be right about every message the attacker chooses to send.
what they saw
Spam reframed from recognition to pricing: identity can't be trusted, filters can't keep up — but computation is real and can't be forged. Price every message in work; bulk abuse stops paying, ordinary use stays free.
the move
They proposed attaching a modest, verifiable computational price to every message: before being accepted, an email must carry a proof that the sender burned a small amount of CPU work to compute it. Generating the proof costs the sender real time; checking it costs the receiver one instant. A person sending a dozen letters a day never notices the price; an operation forging a million a day is priced into bankruptcy — without money changing hands, without micropayment rails, and without anyone ever having to identify the sender.
why it works
The asymmetry is load-bearing: generating the proof is deliberately slow, verifying it is instant, and the cost scales exactly with volume. A human sender pays it once or twice a day without noticing; a mass sender pays it a million times and the price becomes the entire business model of the spam. Because the price is paid in work rather than money, it needs no accounts, no billing and no knowledge of who anybody is — it prices the act itself, which is the one thing an abuser cannot fake.
the payoff
PoW stayed niche against spam — a Microsoft mail-stamp aside — and outlived the mission: via Back's 1997 hashcash it became Bitcoin's engine.
where it breaks
It only works where the honest volume is small and the abusive volume is large — price every use and you tax your own users the moment normal traffic grows heavy. It also fails when the attacker can borrow someone else's compute cheaply, or when the protected action is worth more than the price (the same mechanism later had to be re-priced for Bitcoin because mining the work became profitable). And pricing abuse never replaces removing genuinely harmful content — it only makes the flood uneconomical, not the individual message legitimate.
what came after
The paper sits at the root of proof-of-work as a security primitive: hashcash.org's own record lists Dwork and Naor's CRYPTO '92 paper as the origin of the idea and credits Microsoft's Coordinated Spam Reduction Initiative with a hashcash-style stamp in Exchange, Outlook and Hotmail, while Bitcoin's design builds on the reusable version Back published in 2002. A mechanism invented to make junk mail too expensive to send now orders consensus across thousands of machines — the largest-scale deployment of the trick its inventors proposed for the mailbox.
references
- [1]Pricing via Processing or Combatting Junk Mail (CRYPTO '92)Springer, Lecture Notes in Computer Science 740, 1992link.springer.com
- [2]Hashcash papers and related work — lists Dwork & Naor 1992 as the origin of the proof-of-work ideahashcash.org (Adam Back), 2002hashcash.org
- [3]Bitcoin: A Peer-to-Peer Electronic Cash System — builds the mining mechanism on Back's hashcashbitcoin.org (Satoshi Nakamoto), 2008bitcoin.org