How does the attack success rate change if the number of distributed nodes increase e.g. from 2 to N (where N is the total number of nodes).
Assuming total data poisoned remains the same e.g. if total data points which are poisoned are M, they always get divided equally among all the compromised nodes.
So if the compromised nodes are 2, then each node poisons M/2 data points, whereas if # compromised nodes is k, then each node poisons M/k data points.
As a first step we want to investigate this:
How does the attack success rate change if the number of distributed nodes increase e.g. from 2 to N (where N is the total number of nodes).
Assuming total data poisoned remains the same e.g. if total data points which are poisoned are M, they always get divided equally among all the compromised nodes.
So if the compromised nodes are 2, then each node poisons M/2 data points, whereas if # compromised nodes is k, then each node poisons M/k data points.