None techies, just read the last line for explanation
Lemma 3.0.3. The NXT algorithm described above satisfies the conditions
for being a cryptoeconomically secure entropy source.
Proof. To prove unpredictability, we note that the NXT blockchain produces
a block every minute, and so the update v ← sha256(v, V (β)) takes
place once a minute. During each round of updating, there is a probability
1 − po(60) that the primary signer will be online, and po(60) that the
signer will be offline and thus a secondary signer will need to produce the
block. Hence, after 1
−log(po(60)) blocks, there is a probability p ≈
1
2
that the
resulting value will be the “default value” obtained from updating v with
the primary signers’ public keys at each block, and a p ≈
1
2
probability that
the resulting value will be different. We model 512 iterations of this process
as a tree, with all leaves being probability distributions over sequences
of 512 public keys of signers, where all probability distributions are disjoint
(ie. no sequence appears with probability greater than zero in multiple
leaves). By random-oracle assumption of sha256, we thus know that we have
a set of 2512 independently randomly sampled probability distributions from
{0, 1}
256, and so each value will be selected an expected {0, 1}
256 times, with
standard deviation 2128. Hence, the probability distribution is statistically
indistinguishable from a random distribution.
To show that the first uninfluenceability criterion holds true, note that
the only way to manipulate the result is for the block proposer to disappear,
leading to another proposer taking over. However, this action is costly for
the proposer as the proposer loses a block reward. The optimal strategy
is to disappear with probability 0 < q <= 1 only when the predicate will
be unsatisfied with the proposer participating but will be satisfied with
the next proposer partipating; if a predicate has probability p this entails
disappearing p ∗ (1 − p) ∗ q of the time, meaning that the predicate will be
satisfied p + p ∗ (1 − p) ∗ q of the time instead of p of the time, a probability
increment of p∗(1−p)∗q will have a cost of p∗(1−p)∗q∗R if R is the signing
reward (whose real value is proportional to the quantity of transaction fees, a
reasonable metric of economic activity). Hence, the desired condition holds
true with b = 1.
To show that the second uninfluenceability criterion holds true, note that
when one is not the signer, one has no influence on the entropy, and when
one is the signer one has the ability to not sign and instead defer to the
next signer. Hence, an attacker controlling 1
k
of all signing slots will be able
to defer to the second signer 1
k
of the time, to the third signer 1
k
2 of the
time (by being in the first two slots simultaneously), etc, so in total such an
attacker will on average be able to choose between 1 + 1
k−1
values and thus
multiply the probability of a desired predicate by a factor of 1 + 1
k−1
. If the
attacker controls 1
3
of all signing slots, the result will thus be increasing the
probablity by a factor of 3
2
.
***********
it seems vitalik made a proof about NXT algo