Re:Colonial Pipeline Paid Hackers Nearly $5M in Ransom
by RyanHamilton on Thursday May 13, 2021 @10:35PM
(#27150040)
Great example. It is worth emphasising your model will capture the game from the ransom attackers perspective since for them it is repeated many times. It won't give the EV for the attacked as their number of ransom incidents will hopefully be very small rather than tend to infinity. Which likely means the EV will be based on their own personal utility function.
Re:Colonial Pipeline Paid Hackers Nearly $5M in Ransom
by bumbledraven on Thursday May 13, 2021 @11:35PM
(#27150418)
The Kelly Criterion can be used to analyze the options from the defender’s point of view, e.g.:
• pay for security that prevents attacks = lg[company value – cost of security]
• don’t pay for security = attack probability × lg[company value – cost of dealing with attack] + (1 – attack probability) × lg[company value]
Example: Say you have $10⁵, and you have the option to play game 1, which offers a 95% chance of a $5×10⁶ prize, or game 2, which offers a 50% chance of a $5×10⁷ prize. By the Kelly Criterion, the value of a scenario is the expected value of the logarithm of your wealth under that scenario:
• not playing at all = lg[10⁵] = 5
• game 1 = 95%×lg[10⁵ + 5×10⁶] + 5%×lg[10⁵ + 0] ≈ 6.6
• game 2 = 50%×lg[10⁵ + 5×10⁷] + 50%×lg[10⁵ + 0] ≈ 6.3
So game 1 is the best option if you have only $10⁵. On the other hand, if you have, say, $10⁶, then game 2 is the best option.