Smart sensors can gather all kind of information and process it. Cameras are still dominating and smart cameras can offer services for face recognition or person tracking. To cover a larger area, to save costs and to add more and different sensors, operators build collaborations. Cryptographic methods can achieve integrity and confidentiality between operators, but not trust. Even if a partner or one of his sensors is authenticated, no statements can be made about the authenticity of the sensor data or its quality. Hence, trust must be established between the partners and in their sensors.
Trust can be built based on past experience. A reputation system collects opinions of operators about the behavior of sensors and calculates trust based on these opinions. Many reputation systems have been proposed, e.g., for authentication of files in peer-to-peer networks.
This work presents a new reputation system, which is designed to calculate the trustworthiness of surveillance systems and the authentication of sensor data. A new trust model, including algorithms to calculate and update trust on past experiences, is proposed, as well as protocol for exchanging recommendations. When fusing information of multiple sensors for a surveillance task, it cannot always be reconstructed, which information led to a bad result. Hence, an approach for fair rating is shown. The proposed system has been realized in a Service Oriented Architecture for easy integration in existing surveillance systems. The model itself can be used in every decentralized heterogeneous sensor network.