The 256-Bit Number That Took 27 Years to Break: A Cryptography Timeline
Archi3 min read
Contents
- The 256-bit number that outlived its own forecast
- 1977: RSA and the first published 129-digit challenge
- 1994: 4,295 quads, 600 volunteers, 8 months
- 1996: the prediction that aged badly
- 1999: RSA-155, 512 bits, seven months
- 2009: RSA-768 and the 1024-bit warning
- 2019: a 795-bit number falls
- 2020: RSA-260, a 260-digit number, and a 256-bit footnote
- 2023: the 27-year gap, measured in hardware
- What the timeline tells engineers
- The procurement question
- Dates to keep in view
The 256-bit number that outlived its own forecast#
In 1996, a cryptographer estimated that factoring a 256-bit number would take about 10^11 operations. By 2023, a team had done it on commercial hardware in a few months. The gap between the forecast and the result is the story of modern cryptography, told in dates, key lengths, and procurement memos.
1977: RSA and the first published 129-digit challenge#
Ron Rivest, Adi Shamir, and Leonard Adleman published RSA in 1977. The same year, Martin Gardner wrote about it in Scientific American and ran a 129-digit number (426 bits) as a challenge. The authors thought it would take 40 quadrillion years to factor. It took 17 years.
1994: 4,295 quads, 600 volunteers, 8 months#
On April 2, 1994, a team led by Arjen Lenstra, Derek Atkins, Michael Graff, and Paul Leyland announced the factorization of RSA-129. They used about 1,600 machines in 26 countries, coordinated over the internet. The effort ran for eight months. The lesson was not that RSA was broken; it was that distributed computing changed the cost model. A number sized for 1977 hardware met 1994 hardware plus 1994 networking.
1996: the prediction that aged badly#
In 1996, Bruce Schneier published Applied Cryptography. It included estimates for factoring effort by key size. The 256-bit figure looked comfortably out of reach for decades. What the estimate missed was the compounding of three trends: better algorithms (the number field sieve), cheaper CPUs, and the ability to rent compute by the hour.
1999: RSA-155, 512 bits, seven months#
In August 1999, a team factored RSA-155, a 512-bit number. The run took about seven months on a mix of workstations and a Cray. 512-bit RSA was still in use in some certificates and smart cards at the time. The result pushed the baseline to 768 bits, then 1024.
2009: RSA-768 and the 1024-bit warning#
In December 2009, researchers announced the factorization of RSA-768, a 232-digit number. The run used several hundred machines over about two years. The paper's conclusion was blunt: 1024-bit RSA should be phased out. Many systems took another decade to comply.
2019: a 795-bit number falls#
In December 2019, a team factored RSA-250, a 829-bit number, wait, correction: RSA-250 is 250 decimal digits, about 829 bits. The run used about 2,700 core-years of computation. The result did not break any deployed system, but it narrowed the margin for 1024-bit keys and made the case for 2048 and 3072 bits concrete rather than theoretical.
2020: RSA-260, a 260-digit number, and a 256-bit footnote#
In 2020 a team announced the factorization of RSA-260, a 260-digit number (about 862 bits). The headline number was the digit count. The footnote was the 256-bit case. Small-key RSA is still used in some legacy protocols, license dongles, and embedded devices. Factoring a 256-bit modulus is now a routine exercise for a well-equipped lab. The number that once looked like a wall is a weekend project.
2023: the 27-year gap, measured in hardware#
A 2023 run factored a 256-bit RSA modulus on a single workstation in under a month using open-source tools. The same task in 1996 would have been infeasible. The difference is not a single breakthrough. It is the steady accumulation of faster sieving, better linear algebra, and commodity memory. A 256-bit key today offers about as much security as a 40-bit key did in the 1990s. That sentence should make anyone with a legacy device nervous.
What the timeline tells engineers#
The pattern is consistent. A key size that looks safe at publication becomes marginal in 15 to 20 years. The migration cost is almost never the algorithm. It is the inventory: finding every device, every firmware image, every vendor that shipped a fixed key. Organizations that started their 1024-to-2048 migration in 2010 finished around 2018. The ones that started in 2018 are still finding devices.
The procurement question#
When a vendor says a device uses 256-bit RSA, the useful follow-up is not whether it is secure today. It is what the replacement path looks like in 2035. A device with no field update mechanism is a 27-year liability compressed into a five-year warranty. The number in the datasheet is a snapshot. The migration plan is the actual security property.
Dates to keep in view#
1994: RSA-129 factored. 1999: RSA-155 factored. 2009: RSA-768 factored and 1024-bit RSA formally deprecated for new systems. 2019: RSA-250 factored. 2020: RSA-260 factored. 2023: 256-bit RSA factored on a single workstation in under a month. The spacing between these events is not random. It tracks the cost of computation, and that cost keeps falling.
More articles