The main aim of this master's thesis is to implement a common, secure and peer-to-peer communication system. The system has ability to automatically establish and run a secure end-to-end connection. It has this ability even if a network address translator is in the way to the destination system, without need of any explicit configuration of this translator. The security procedures of this system are in a transparent manner masked from individual applications, which had to solve this challenge in their own way. A responsibility for a security is delegate to an application-independent subsystem working within the core of an operating system. The security of this subsystem is based on capturing the outbound and inbound IP packets and their authentication and encryption. The system was successfully implemented in MS Windows XP operating system, in programming language C++. Transfer rate of communication tunnel in different network bandwidth speeds was measured. Result shows, that in the case of use the system on standard PC sold nowadays is practically no decrease of the transfer rate in comparison to a common channel.

The seamless retrieval of information distributed across networks has been one of the key goals of many systems. Early solutions involved the use of single static agents which would retrieve the unfiltered data and then process it. However, this was deemed costly and inefficient in terms of the bandwidth since complete files need to be downloaded when only a single value is often all that is required. As a result, mobile agents were developed to filter the data in situ before returning it to the user. However, mobile agents have their own associated problems, namely security and control. The Agent Trees Multi-Agent System (AT-MAS) has been developed to provide the remote processing and filtering capabilities but without the need for mobile code. It is implemented as a Peer to Peer (P2P) network of static intelligent cooperating agents, each of which control one or more data sources. This dissertation describes the two key technologies have directly influenced the design of ATMAS, Peer-to-Peer (P2P) systems and Multi-Agent Systems (MAS). P2P systems are conceptually simple, but limited in power, whereas MAS are significantly more complex but correspondingly more powerful. The resulting system exhibits the power of traditional MAS systems while retaining the simplicity of P2P systems. The dissertation describes the system in detail and analyses its performance.