Security in Ambient Intelligence (AmI) poses too many challenges due to the inherently insecure nature of wireless sensor nodes. However, there are two characteristics of these environments that can be used effectively to prevent, detect, and confine attacks: redundancy and continuous adaptation. In this article we propose a global strategy and a system architecture to cope with security issues in AmI applications at different levels. Unlike in previous approaches, we assume an individual wireless node is vulnerable.
We present an agent-based architecture with supporting services that is proven to be adequate to detect and confine common attacks. Decisions at different levels are supported by a trust-based framework with good and bad reputation feedback while maintaining resistance to bad-mouthing attacks. We also propose a set of services that can be used to handle identification, authentication, and authorization in intelligent ambients.
The resulting approach takes into account practical issues, such as resource limitation, bandwidth optimization, and scalability.