This thesis focused on Sybil attack analysis in Peer-to-Peer botnet based on Kademlia network protocol and architecture. With the consideration of the node space may not be full, each node location was assumed to have a probability p to be taken. The objective of this research is to theoretically evaluate such botnet/network’s robustness under two types of malicious attack - random Sybil attack and fake target attack. Probability of successfully finding a target node that can be possibly locating anywhere in the network is taken as the main evaluation factor. With different key parameters, mathematical models are established and partial closed-form formulas have been proposed. Numerical results generated by Matlab are demonstrated and analyzed to conclusion.