Quote from: TheFascistMind on October 03, 2014, 12:33:46 AM

Quote from: TheFascistMind on October 02, 2014, 10:53:53 AM

Did you miss the entire discussion about permutations of consecutive independent trials (i.e. not separated by 65 minutes each)?

Quote from: JorgeStolfi on October 02, 2014, 02:10:48 PM

If someone is causing the block rate to be higher than one per minute, that should be detected by counting blocks in some long interval (say, 10 hours) .

Afaics, that won't help you identify an intentional segregation of fast and slow blocks to manipulate the 80/20 discard window of the CN difficulty adjustment algorithm.

Quote from: JorgeStolfi on October 02, 2014, 02:10:48 PM

If the block rate is OK but the suspicion is that the timing of blocks is being manipulated, that should be detected by plotting a histogram of block-to-block gaps, or of number of blocks in successive 2 minute intervals, again over a long enough period.

I don't see how that will identify an intentional segregation since the 80/20 discard is relative to its own statistics? Do you mean comparing histogram histories?

Quote from: JorgeStolfi on October 02, 2014, 02:10:48 PM

Computing the probability of a certain complicated pattern occurring, after seeing it occur, is a tricky business. The chance of my mother marrying my father was one in two billions or so; that does not mean that my mere existence is a sign that something fishy is going one with the universe...

You said you read the upthread discussion, yet you continue the strawman. My point was to refute the anti-FUD-campaign which was turning into a Monica Lewinsky or Steve Jobs denial, "no malfunction in our devices"[1].

[1] "don't touch it that way"

Sorry, I know practically nothing about the Monero protocol, so I cannot say anything useful about the "attack" specifically. (The continuous difficulty adjustment and the 20% outlier rejection filter seem to make it hard to model statistically. If the difficulty gets adjusted diring the data collection interval, one cannot assume that block finding is a Poisson process; unless the event times are remapped to a suitable variable-rate clock.)

I was only commenting on the suggestion (perhaps not even by you, it is hard to keep track of the debate) that the occurence of a pattern that has very low probability of occurring is evidence of manipulation. It may be evidence, if the probability analysis is properly done, but it is very easy to slip and see manipulation where there isn't.

The mistake is easy to make if one takes a complicated pattern that has occurred. Others have pointed out that fallacy. If the pattern covers a dozen consecutive events, its probability will be very low -- but some pattern must occur at every point,so nothing strange there.