random numbers certified by bells theorem
DESCRIPTION
A brief talk about a paper published by a bunch of people in 2010 on how it is possible to use create devices for generating random numbers in such a manner that anyone can verify that a pseudo-random number generator is not being used, but instead that quantum noise is being used.TRANSCRIPT
Random Numbers Certified by Bell's Theorem by Pironio and 10 other authors
from 6 different universities and institutions.
A brief introduction by David Kemp
Licensed under a Creative Commons Attribution 3.0 Licensehttp://creativecommons.org/licenses/by/3.0/
Why?
Pseudo random numbers may have patterns known to attackers.
There is need for a source of randomness that can be verified to be genuinely random.
Bell’s Inequality
• Published by John Stewart Bell in 1964• Proved “reality not separable from observation”.• Einstein died in 1955 believing reality must be
separable from observation (see EPR paradox).
Simple Inequality Example
Room full of people:•Only one right-handed male.•Only one blue-eyed female.•How many (max) blue-eyed right-handers?– At most one of them can be male.– At most one of them can be female.– Hence a maximum of 2.
Simple Inequality as Venn Diagram
Female
Right-handed
Blue-eyed
Size(blue-eyed right-handers) ≤ Size(blue-eyed females) + Size(right-handed males)
In Probabilistic Terms
Pr(blue-eyed right-hander) ≤Pr(blue-eyed female) + Pr(right-handed male)
What if your data is incomplete?
• Suppose you can only ask each person one of:– Are you a blue-eyed right-hander?– Are you a blue-eyed female?– Are you a right-handed male?
What if your data is incomplete?
•Suppose you can only ask each person one of:–Are you a blue-eyed right-hander?–Are you a blue-eyed female?–Are you a right-handed male?
Pr(blue-eyed right-hander) ≤Pr(blue-eyed female) + Pr(right-handed male)
ONLY IF THEY DO NOT LIE!
Identical Twins
• Second room with a twin of everyone in first.• No communication between the rooms.• Ask every person only one of these questions:– Are you blue-eyed?– Are you left-handed?– Are you female?
• Sometimes ask both twins the same question.• Only way to avoid lie detection is prior agreement.
Bell’s Inequality
IF
•A, B and C are binary properties,•A', B' and C' are properties of twin,• Pr(x & not(x')) = 0.
THEN
Pr(A & not(B')) + Pr(B & not(C')) ≥ Pr(A & not(C'))
Pairs of Magic Boxes
A B C A B C
Three buttons and two lights on each box
Magic Box PairsA B C
Can only press one button on each box.
A B C
Magic Box Properties
Whenever same button is pressedon both boxes:
Same light flashes
A B C A B C
•Pr(A:red & B:yellow) ≈ 7.5%•Pr(B:red & C:yellow) ≈ 7.5 %•Pr(A:red & C:yellow) ≈ 25%
Pr(A:red & B:yellow) + Pr(B:red & C:yellow)< Pr(A:red & C:yellow)
A B C A B CMagic Box Properties
But surely that is impossible!!!
Image by Wild Guru Larry. Some rights reserved.
http://www.flickr.com/photos/wentzelepsy/4435803492/
Relies on Quantum Entanglement
Entangled Electrons
• Measure Electron Spin Direction• Spin-up/down instead of clockwise/anti-clockwise• Result seems to be completely random.• Can measure from different angles.–Eg. Vertically, horizontally, and 45º to horizon.
• Measurement disturbs the electron.–Cannot reliably measure same electron from different angles.
Entangled Electrons
Let buttons A, B and C, measure at 0º, 45º and 90º•Spin-up = red•Spin-down = yellow
Entangled Electrons
Measure entangled electrons from same direction:
100% Correlation
(Press same button on both boxes and lights agree)
Entangled Electrons
• A:red & C:red 25%• A:yellow& C:yellow 25%• A:red & C:yellow 25%• A:yellow & C:red 25%
Measure entangled electrons at 90º to each other:50% Correlation
Pr(A:red & B:yellow) + Pr(B:red & C:yellow)< Pr(A:red & C:yellow)
Entangled Electrons
• A:red & B:red ≈ 42.5%• A:yellow& B:yellow ≈ 42.5%• A:red & B:yellow ≈ 7.5%• A:yellow & B:red ≈ 7.5%
Measure entangled electrons at 45º to each other:≈ 85% Correlation
Pr(A:red & B:yellow) + Pr(B:red & C:yellow)< Pr(A:red & C:yellow)
Entangled Electrons
• B:red & C:red ≈ 42.5%• B:yellow& C:yellow ≈ 42.5%• B:red & C:yellow ≈ 7.5%• B:yellow & C:red ≈ 7.5%
Measure entangled electrons at 45º to each other:≈ 85% Correlation
Pr(A:red & B:yellow) + Pr(B:red & C:yellow)< Pr(A:red & C:yellow)
Quantum Entanglement is the only known process that
can violate Bell’s Inequality.
Verifiably Random
We can use these magic boxes to
generate verifiably random numbers
Selecting buttons A, B and C.
How do you select buttons A, B and C?Need initial private random seed.
Does not need to be strongly random!
Final Step
Result may not be uniformly random.
Use known randomness extraction techniques.
Practicalities
• Actually only need two buttons on each box (using the CHSH inequality).
• Difficult to prevent quantum entanglement quickly degenerating.
Caveats
• Need to prevent signaling between the boxes.• This “protocol is not yet proven to be universally-
composable against a full quantum adversary”.
References
•Pironio et al. Random Numbers Certied by Bell's Theorem, Nature v.464, p.1021 (2010).
•J. S. Bell, On the Einstein Podolsky Rosen Paradox, Physics 1, 3, 195-200 (1964)
•A. Einstein, B. Podolsky, and N. Rosen, Can Quantum-Mechanical Description of Physical Reality Be Considered Complete? Phys. Rev. 47, 777–780 (1935)
Schrödinger's Cat