Lukas Klejnowski

Towards Trust in Desktop Grid Systems

The Organic Computing (OC) Initiative deals with technical systems, that consist of a large number of distributed and highly interconnected subsystems. In such systems, it is impossible for a designer to foresee all possible system configurations and to plan an appropriate system behaviour completely at design time. The aim is to endow such technical systems with the so-called self-X properties, such as self-organisation, self-configuration or self-healing. In such dynamic systems, trust is an important prerequisite to enable the usage of Organic Computing systems and algorithms in market-ready products in the future. The OC-Trust project aims at introducing trust mechanisms to improve and assure the interoperability of subsystems. In this paper, we deal with aspects of organic systems regarding trustworthiness on the subsystem level (agents) in a desktop grid system. We develop an agent-based simulation of a desktop grid to show, that the introduction of trust concepts improves the systems performance, in such that they speed up the processes on the agent level. Specifically, we investigate a bottom-up self-organised development of trust structures that create coalition groups of agents that work more efficiently than standard algorithms. Here, an agent can determine individually to what extent it belongs to a Trusted Community.

Efficiency and Robustness Using Trusted Communities in a Trusted Desktop Grid

Bringing forward trust from social systems to system-to-system level can lead to efficiency and robustness improvements in self-organising complex systems. In this paper, we show how trust can enhance the matchmaking and robustness regarding malicious nodes in a Desktop Grid and Volunteer Computing System (DGVCS). We give the agent a suited degree of autonomy to continuously adapt to their environment in both worker and submitter role. In this paper, we introduce adaptivity in submitter and worker role for our trust-adaptive agents. The evaluation shows that using trust based adaptivity algorithms in a DGVCS leads to efficiency improvements and robustness regarding malicious nodes.

An Architecture for Trust-Adaptive Agents

This paper describes a generic architecture for agents that allows agents to adapt to dynamic environments by using a trust mechanism in conjunction with machine learning. Central parts of the architecture are an Observer/Controller loop and a trust aggregation mechanism. It is shown how this proposed architecture can be applied to agents that act on behalf of a user in a decentralised desktop grid computing system. Furthermore, first results of an evaluation in this scenario are discussed.

Formal Specification and Analysis of Trusted Communities

Trusted Communities are a way to improve the performance of self-organising agent systems by limiting the interactions to trustworthy partners and excluding agents that behaved uncooperatively. We describe the mechanism in an abstract way and identify three central decisions each agent has to make if it supports Trusted Communities. Based on a formal specification of the agent behaviour in an instantiation of the mechanism for Desktop Grid Systems, we identify one of the system goals and show by formal verification that this goal is reached. Additionally, we provide certain requirements for the decision procedures that become evident during the analysis process.

Using Trusted Communities to Improve the Speedup of Agents in a Desktop Grid System

In open, technical, multi-agent based systems, self-interested agents can show behaviours that degrade the performance of the system. This can be countered by providing cooperation incentives. In this paper, we present a formalisation of delegation incentives for an open, agent-based, Desktop Grid system, based on decision trees. We then discuss reputation-based delegation strategies, as well as replication-based delegations strategies and focus on the incentives these strategies provide for agents to cooperate. We further show why we see room for improvement, and how this can be achieved with organisation-based delegation. We propose a delegation strategy based on Trusted Communities and present evaluation results for the comparison of these strategies with respect to the achieved average speedup in the system.

Interactive Graph View of Explicit Trusted Communities in an Open Trusted Desktop Grid System

We present self-organisation processes within a decentralised P2P open Desktop Grid System. Every agent aims to maximise the speedup of his jobs by joining the system. Therefore he needs to cooperate with other agents. We use a trust metric to assess the willingness of cooperation for other agents. To further optimise their speedup agents can form explicit Trusted Communities with trusted peers. Thus they improve their performance and system robustness. The demos show a visual representation combining global effects and local interactions in an interactive graph view.

Trust Communities: An Open, Self-Organised Social Infrastructure of Autonomous Agents

Future technical systems will be increasingly characterised by openness and heterogeneity of participating elements. Based on exemplary application scenarios such as Desktop Computing Grids, Smart Power Grids, and Networked Camera Systems, this chapter develops a solution perspective to handle anomalies, disturbances, and malicious behaviour by making use of trust and reliability measures in self-organised systems. The overall goal is to increase the robustness of open distributed systems with low overhead. Therefore, a novel self-organised multi-agent organisation—the Trust Community—is introduced in two variants: as implicit and as explicit self-structuring society of autonomous agents. For the explicit variant, a life-cycle and management routines are described. For evaluation purposes, we simulate a Trusted Desktop Computing Grid and introduce different types of stereo-type agent behaviour, ranging from altruistic to egoistic and to cunning behaviour. In order to support efficiency and stabilise the process, we show the benefits of explicit Trust Communities, which results in significantly lower overhead and more reliable relations among agents compared to other forms of agent societies.

Advanced Attacks to Trusted Communities in Multi-agent Systems

Self-integration of system components is characterised by uncertain relations among these components. One approach to handle such uncertainties is to introduce technical trust. In previous work, we developed a concept to automatically build collections of components with stable trust relationships so-called Trusted Communities. In this paper, we discuss advanced attacks to the trusted communities and describe counter measures to deal with malicious elements. As application scenario, a Desktop Grid Computing system is used, since it mischaracterised by all those properties that become important when developing self-integrating systems: e.g. openness, heterogeneity and limited control over participating entities. The evaluation demonstrates that even advanced attacks can be recognised fast and successfully.

Estimation of Reward and Decision Making for Trust-Adaptive Agents in Normative Environments

In an open trusted Desktop Grid system with a normative environment incentives and sanctions may change during runtime. Every agent in the system computes work for other agents and also submits jobs to other agents. It has to decide for which agents it wants to work and to which agent it wants to give its jobs. We introduced a trust metric to isolate misbehaving agents. After getting a job processed by another agent it will get a reward. When processing a job for another agent it will get a positive trust-rating, but no direct reward. To come to a decision when accepting or rejecting jobs we need to be able to estimate the reward. Since the environment may change at runtime and to overcome delayed reward issues we use a neural network to estimate the reward based on the environment and trust level.

Self-Organisation and Evolution for Trust-Adaptive Grid Computing Agents

The Organic Computing (OC) initiative aims at introducing new, self-organising algorithms in order to cope better with the complexity of today’s systems. One approach to self-organisation is the introduction of agents which are able to continuously adapt their behaviour to changing environmental conditions and thus collectively create an efficient and robust system. In this paper, we introduce an evolutionary approach to an agent which acts autonomously and optimises its behaviour at run-time. The behaviour of the Evolutionary Agent is defined by ten chromosomes. When two agents interact, the inferior agent copies a part of the genes of the more successful agent. Therefore, the most successful gene combination will spread throughout the network. Application scenario for our evaluation is the Trusted Desktop Grid, a distributed system where computing resources are shared by autonomously acting agents.