The Internet of Things, i.e., millions to billions of low-end smart devices connected to the Internet, is becoming a reality. There have been important research initiatives addressing the problem of defining reference models and new alternative communication architectures. However, there have been comparatively little efforts dedicated to the definition of platforms for the deployment and execution of Machine-to-Machine (M2M) applications, which are the real drivers of innovation. In fact, M2M applications reduce the gap between the real and virtual world, and allow the secure and dependable automation of tasks to improve industrial productivity and quality of life of citizens.Currently, M2M applications are assumed to reside within systems in data centres, i.e., within the "cloud". However, we argue that such a design is sub-optimal for M2M because of its inherent characteristics: typical cloud applications require data to be available at any location and for very long periods for use by human begins, whereas M2M applications have a very limited scope in time and space, require simple and repetitive interactions, and must respond in a consistent manner to avoid service disruption.In BETaaS we propose a platform for the execution of M2M applications, which is built on top of services deployed in a "local cloud" of gateways, the latter being the devices which provide the smart things with connectivity to the Internet (e.g., smart phones, home routers, road-side units). Adaptation layers will be defined to interconnect BETaaS with the main architectures proposed at a European level for M2M communication, including ETSI M2M and IoT-A. In addition to being highly scalable, security, dependability, and QoS features will be embedded "by design" into the BETaaS platform, which will be released as open source to the community. Platform validation will be done through experiments in two testbed facilities, targeting the Smart City and Home Automation use cases.