Designing a Network Infrastructure for Survivability of Multi-Agent Systems (original) (raw)
Related papers
Agent-oriented design for network survivability
5th International Conference on Intelligent Systems Design and Applications (ISDA'05), 2005
Intelligent behavior is the selection of actions based on knowledge. The design of the Fuzzy Adaptive Survivability Tool (FAST) agents and their intelligent behavior is explained. A FAST agent uses Belief-Desire-Intention (BDI) logic as the reasoning framework to decide on desirable response plans. These decisions are both context-sensitive to take into account the changes in the network status and costsensitive to avoid the risk of collateral damage. A realworld scenario, which shows how the FAST agents choose desirable responses to mitigate scanning worm traffic, is also presented.
2020 IEEE Conference on Control Technology and Applications (CCTA)
This paper aims at investigating a novel type of cyber attack that is injected to multi-agent systems (MAS) having an underlying directed graph. The cyber attack, which is designated as the controllability attack, is injected by the malicious adversary into the communication links among the agents. The adversary, leveraging the compromised communication links disguises the cyber attack signals and attempts to take control over the entire network of MAS. The adversary aims at achieving this by directly attacking only a subset of the multi-agents. Conditions under which the malicious hacker has control over the entire MAS network are provided. Two notions of security controllability indices are proposed and developed. These notions are utilized as metrics to evaluate the controllability that each agent provides to the adversary for executing the malicious cyber attack. Furthermore, the possibility of introducing zero dynamics cyber attacks on the MAS through compromising the communication links is also investigated. Finally, an illustrative numerical example is provided to demonstrate the effectiveness of our proposed methods.
Distributed Algorithms for Dynamic Survivability of Multiagent Systems
Lecture Notes in Computer Science, 2004
Though multiagent systems (MASs) are being increasingly used, few methods exist to ensure survivability of MASs. All existing methods suffer from two flaws. First, a centralized survivability algorithm (CSA) ensures survivability of the MAS-unfortunately, if the node on which the CSA exists goes down, the survivability of the MAS is questionable. Second, no mechanism exists to change how the MAS is deployed when external factors trigger a re-evaluation of the survivability of the MAS. In this paper, we present three algorithms to address these two important problems. Our algorithms can be built on top of any CSA. Our algorithms are completely distributed and can handle external triggers to compute a new deployment. We report on experiments assessing the efficiency of these algorithms.
Survivability of a distributed multi-agent application - a performance control perspective
2005
Distributed multi-agent systems (DMAS) such as supply chains functioning in highly dynamic environments need to achieve maximum overall utility during operation. The utility from maintaining performance is an important component of their survivability. This utility is often met by identifying trade-offs between quality of service and performance. To adaptively choose the operational settings for better utility, we propose an autonomous and scalable queueing theory based methodology to control the performance of a hierarchical network of distributed agents. By formulating the MAS as an open queueing network with multiple classes of traffic we evaluate the performance and subsequently the utility, from which we identify the control alternative for a localized, multi-tier zone. When the problem scales, another larger queueing network could be composed using zones as building blocks. This method advocates the systematic specification of the DMAS's attributes to aid realtime translation of the DMAS into a queueing network. We prototype our framework in Cougaar and verify our results.
Building a Distributed Network using Agent Migration 1
2012
The efficient utilization of network resources is an important issue. The problem is hard due to the distributed nature of computer networks, high communication demands and the desire for limited communication overheads. One solution to such challenges is to design efficient, decentralized and fault-tolerant data propagation model which accomplishes tasks with no access to global network information. Mechanisms of agent propagation are useful because agents can be organised into efficient configurations without imposing external centralised controls. Operation without a central coordinator eliminates possible bottlenecks in terms of scalability and reliability
Constrained coalitional games and networks of autonomous agents
2008 3rd International Symposium on Communications, Control and Signal Processing, 2008
We develop a unifying analytical and optimizationbased framework for the design, operation and performance evaluation of networks of dynamic autonomous agents. The fundamental view is that agents in such a network are dynamic entities that collaborate because via collaboration they can accomplish objectives and goals much better than working alone, or even accomplish objectives that they cannot achieve alone at all. Yet the benefits derived from such collaboration require some costs (e.g. communications), or equivalently, the collaboration is subject to constraints. Understanding and quantifying this tradeoff between the benefits vs the costs of collaboration, leads to new methods that can be used to analyze, design and control/operate networks of agents. Although the inspiration for the framework comes from social and economic networks, the fundamental ideas and in particular the methodology of dynamic constrained coalitional games (DCCG) can unify many concepts and algorithms for networks in various areas: social networks, communication networks, sensor networks, economic networks, biological networks, physics networks. We then analyze a specific instance of such tradeoffs arising in the design of security aware network protocols. We extend network utility maximization (NUM) so as to encompass security metrics such as "trust". The trust values assigned to nodes are based on interaction history and community-based monitoring. The effect of these trust values on the resulting protocols is that in routing and media access scheduling node trustworthiness is automatically considered and used. We develop a distributed algorithm for the joint physical-MAC-routing protocol design. Our extension to NUM with security concerns leads to resilient networks.
Dependency of Network Structures in Agent Selection and Deployment
2006 IEEE/WIC/ACM International Conference on Intelligent Agent Technology, 2006
This paper shows that the statistical properties of the network topology are indispensable information for improving performance of multi-agent systems (MASs), though they have not received much attention in previous MAS research. In particular we focus on the applicability of the degree of an agent-the number of links among neighboring agentsto load-balancing for the agent selection and deployment problem. The proposed selection algorithm does not need global information about the network structure and only requires the degree of a server agent and the degrees of the nodes neighboring the server agent. Through simulation of several topologies reproduced by the theoretical network models, we show that the use of the local topological information significantly improves the fairness of the servers even for a large-scale network. We also find that the key mechanisms for load-balancing in a given network topology are highly asymmetric degree characteristics (scalefree) and the negative degree correlation.
Service for fault tolerance in the Ad Hoc Networks based on Multi Agent Systems ∗
2014
The Ad hoc networks are distributed networks, self-organized and does not require infrastructure. In such network, mobile infrastructures are subject of disconnections. This situation may concern a voluntary or involuntary disconnection of nodes caused by the high mobility in the Ad hoc network. In these problems we are trying through this work to contribute to solving these problems in order to ensure continuous service by proposing our service for faults tolerance based on Multi Agent Systems (MAS), which predict a problem and decision making in relation to critical nodes. Our work contributes to study the prediction of voluntary and involuntary disconnections in the Ad hoc network; therefore we propose our service for faults tolerance that allows for effective distribution of information in the Network by selecting some objects of the network to be duplicates of information.
Distribution of mobile agents in vulnerable networks
Concurrency and Computation: Practice and Experience, 2007
Advances in the Internet and the computer industry have created many new application areas for network routing such as Grid computing and also brings new challenges to traditional routing techniques. In this paper we propose a mobile agent-based routing model in vulnerable networks for these applications. To characterize the behaviors of mobile agents and their effects on the network performance, we analyze the population distribution of mobile agents as a measurement of the computational resource consumption. Our analysis reveals theoretical insights into the statistical behaviors of mobile agents and provides useful tools for effectively managing mobile agents in large networks.
Resource allocation for multi-agent problems in the design of future communication networks
2007
new architecture for a smarter Internet abstracts the functional components of the network from the hardware that enables it. This is done via software agents that implement such functions and are capable of relocating themselves over the network to optimize resources. To achieve that we need an algorithm that governs the way agents distribute themselves in the physical network, and this paper presents our first effort towards this goal. We formulate the problem as an optimization problem, where agents must be distributed in the network, and can receive resources from the nodes they occupy according to their task requirements. This optimization problem is then solved in a hierarchical manner: A centralized (randomized) algorithm optimizes the agent distribution among the nodes, and a decentralized (convex optimization) algorithm performs the resource allocation within each node. We present simulation results showing that the hierarchical optimization algorithm achieves the desired objective. Finally we discuss our next steps towards decentralizing the proposed algorithm.