A Cloud Model Based Trust Evaluation Model for Defense Agent

All defense agents (DAs) are trustworthy and controllable by default during the implementation of defense scheme in the computer network collaborative defense (CNCD) system. But this unreasonable assumption does not hold in the open network environment. Malicious agent will be led into the deployment of CNCD defense scheme and the fail rate of defense schemes will be raised under this assumption, which will decrease the security of the whole system. To address this issue, trust evaluation should be conducted. In the present research work, a trust evaluation model of CNCD is proposed. The model can describe trust from the aspects of randomness and fuzziness, and conduct trust updating. The trust evaluation model includes two key parts: task execution evaluation and defense agent trust updating. Evaluation functions of DAs’ feedback, including functions of finish time (FT) and defense quality (DQ), are studied in detail. Two properties of trust, including time decay and asymmetry, are adopted in the evaluation functions of DAs’ feedback. A sliding time window-based dual weight direct trust cloud model (STBCM) is likewise proposed for trust updating. The contrast experiments show that the proposed algorithm has lower fail rate of defense scheme, and can provide support for the trust deployment of the CNCD scheme.