Research On Survivability Evaluation Technology Towards Wireless Sensor Networks

Survivability refers to the ability that one system can continue providing critical services under attack or fault.The feature that wireless sensor networks are vulnerable to attack and fault makes it very important to study its survivability.Therefore,in order to evaluate the survivability of WSN,it is necessary to establish specific and accurate survivability models and design reasonable indicators.There are two shortcomings in existing models and indicators: First,the survivability model cannot accurately reflect the state changes of nodes and networks;second,most of existing indicators can only evaluate the long-term survivability of networks,and cannot dynamically reflect the survivability in a certain moment or under a certain period of time.To solve these problems,this thesis first proposes a single-node survivability model based on CTMC,which reflects the state change of a single node under attack accurately and reasonably,then designs and implements two survivability indicators: single-node attack repeatability and quantity of single-node attack.Further,the cluster-based combined CTMC model is proposed.By introducing a threshold mechanism,the networks' overall survivability can be characterized more accurately,also quantity of transient attack and quantity of continuous attack are designed and implemented.The main work of this thesis is as follows:1.Establish a single-node survivability model,design evaluation indicators and implement measure algorithms.According to the characteristics of a single node under attack,a state transition model based on CTMC is established to describe the behavior of a single node.Two evaluation indicators are designed to evaluate single-node survivability: 1)single-node attack repeatability,which means the probability that a single node will experience repeated attacks in a certain period of time;2)quantity of single-node attack,which means the number of attacks that occur when a node reaches a certain moment.According to the characteristic that the CTMC model is able to calculate continuous time probability,the single-node attack repeatability algorithm is realized;by introducing the Markov model with reward structure,the quantity of singlenode attack algorithm is implemented.Experiments show that the estimated value obtained by the index algorithm is consistent with the trend of actual value obtained by the existing DoS attack detection method,which means our algorithm is reasonable.The influence of the transfer rate of different states is also investigated,the greater the rate of entering the attack state,the higher the repeatability of the attack and the quantity of attack,and the improvement of the response rate can significantly reduce the repetitiveness of the attack.The single node attack quantity increases with time.2.Establish a network survivability model,design evaluation indicators and implement measure algorithms.Based on the single-node model,a cluster-based combined CTMC model is proposed to characterize the evolution of the network's overall state of survival.Firstly,a threshold mechanism is introduced to trigger the state transition of a single cluster,based on this,a combined CTMC model between clusters is proposed to describe the overall state of survivability of the network,solving the problem of inaccurate network characterization.To evaluate network survivability,two evaluation indicators are designed: 1)quantity of continuous attack,which means the average number of nodes that the network as a whole was attacked for a certain period of time;2)quantity of transient attack,which means the average number of nodes that the network as a whole was attacked at a certain time.The CTMC model is used to implement the quantity of continuous attack algorithm because it can calculate the state transition probability at continuous time;through the conversion between some moment and the round of network,the uniformed DTMC model is used to implement the quantity of transient attack algorithm.By comparing with the actual quantity obtained by existing Sybil attack detection scheme,it shows that the estimated value obtained by the index algorithm matches the change trend of the actual value,which means our algorithm is reasonable.Experiments also show that the continuous attack quantity and transient attack quantity will increase with time.Increasing the attack response rate can reduce the continuous attack quantity and the transient attack quantity will maintain an increase under any circumstances.