Silke Hoeppner, Matthias Welwarsky, Brendan Murphy TOWARDS AGENT-PLATFORM INDEPENDENT MIGRATION OF MOBILE AGENTS
Abstract: This paper proposes the concept of an interface between mobile agent platforms, enabling agents to migrate between proprietary platforms from different vendors.
Keywords: Mobile agents, FIPA, OMG, (legacy) agent platforms, proxies
An agent is a self-contained sociable software or hard/software system, which exists in a predefined environment where it posesses the ability for autonomous and flexible (reactive and pro-active) actions. Mobile agents are agents that are able to roam the network in behalf of their users, trying to fulfil certain tasks like configuring remote systems, buying tickets etc. In order to live on a machine, the agent needs to have a virtual environment, which enables it to act on its own behalf. This is the agent platform.
The emergence of proprietary agent platforms lead to a community of isolated solutions, which may work sufficiently in their physical and logical domains, but which lacked every possibility to enable agent communication or even migration between differently constructed platforms. The underlying conceptional problems used to be the lack of standards in communication between agent platforms. This was the motivation for the FIPA group to specify a standard, which should ensure communication between agents on different platforms as well as communication between platforms themselves. So FIPA specifies services that have to be fulfilled by platforms, in order to enable agents from FIPA-conformant platforms, which may have been developed by different vendors, can communicate.
However, when it comes to migration, the following problems still arise: Each platform provides its own interface to agent mobility. Even if platforms are FIPA-compliant, it is still not transparent, what base classes have to be supported in order to ensure that an agent which migrates in between agent platforms, is fully functional on the FIPA compliant target platform. In order to overcome this problem, a proposal to create a standardized method-based interface has been suggested by members of the OMG.
However, even if we imagine a world-wide network of FIPA-compliant agent platforms, there still exists a community of legacy agent platforms, which are not FIPA-compliant but otherwise established. A mobile agent roaming the internet will not be able to migrate to these legacy platforms and may therefore stay unable to reach certain required resources. The solution we propose to this problem is a host, which holds an interface core, which is plugged in between the corresponding agent platforms.
Figure 1: Basic concept of the interface
Our approach to solve the problem of inter platform migration is to have the agent consist of an application specific part that implements the user-specified characteristics of the agent and a set of Java classes that deal with basic functionalities like mobility, message passing and event handling. These base classes implement an interface between the user-specified characteristics and the platform specific (i.e. vendor specific) services. The agent itself becomes vendor independent. The base classes can therefore be separated from the application specific classes and may be exchanged with a set of classes specific to another vendor, as long as the same, standardised interface is implemented.
The notion of this proposal is described on the base of connecting the platforms Concordia from Mitsubishi Electric and Object Space’s Voyager.
This paper also deals with the consequences of the proposed interface concerning security and gives an outlook on the potential and the risks of applying this new technology.
Ingo Stengel, Silke Hoeppner, Brendan Murphy, Aidan McDonald SECURITY IN INTELLIGENT MOBILE AGENT SYSTEMS
Abstract: This paper gives an evaluated state of the art on security methods applied to intelligent mobile agent systems as well as their potential and the following consequences.
Since the invention of mobile agents, researchers have wondered about the killer application for this technology. The main reason for the difficulties in finding such an application is the lack of intelligence and autonomy which mobile agents generally suffer. But in order to apply intelligent mobile agents, that can roam autonomously the intra- and internet, a network of FIPA-compliant and OMG-MASIF–oriented agent platforms is needed. The employment of such systems in the commercial world, i.e. as information brokering agents, depends on a reliable security concept with well developed security policies. The major aspects of the security issue in intelligent mobile agent systems are based on confidentiality, integrity, authentication and non-repudiation. Based on the malicious host problem, in each of these areas, still a lot of security problems exist. Just a few of the rising problems can be solved using classical methods. Therefore mobile agent security requires new, non-classical approaches, which can meet the requirements of the new challenge. The above mentioned problems can be partially solved, i.e. by using encrypted functions that need not to be decrypted for invocation, or using self-changing code which veils its real functionality. This paper gives an evaluated survey of security relevant methods which, applied to intelligent mobile agent systems, help to satisfy the security requirements of such a distributed system.
O. I. Pyatkovsky, K. A. Levin ADAPTIVE HUMAN-MACHINE INTERACTION METHODS IN INTELLECTUAL INFORMATION SYSTEMS
This work suggests using artificial intellect technologies to organize intellectual interfaces with program-technical complexes. Methods of system adaptation to a concrete specialist have been developed. For this human actions recognizing and information system current state functions are realized as well as user class definition and, depending on this, organizing his continuos education in compliance with relevant educational level. The report deals with the approach of organizing user's intellectual interfaces on the basis of initial documents entering program system in information system of enterprise financial and operational activities' accounting and analysis.
The following functions are realizes in the system: teaching the user to work with the system, user's qualification assessment, setting up educational tests when necessary, control of user's progress in system studying in the process of activities. While data input intellectual system operating two levels of control over the input information are applied. Data and user's actions logical control is used on the first level. For this an Expert System (ES) has been developed, which realizes reasoning on the basis of precedent theory . While entering the data into information fields, user actions' system control is applied on the basis of the projected knowledge base. As a result of ES operation control messages and generalizing parameters showing the user's class are formed. After inputting "n" records, their average indication is calculated. Evaluation scale and "n" number are defined by experimental studies method and expert quantitative assessment. Received parameters are used on the second level of data logical control. For this an educated neuron system (NS) is used . Network operation output parameter is the user's class. It defines the user's qualification and work quality criterion. Depending on these parameters' indication automatic choice of educational test and input program adjustment take place. Educational instructions are shown on the video-terminal monitor.
Developed complex of document input intellectual programs with a built-in expert system realizing the rules of precedents and neuron-imitator is included into financial-analytical system "Analyst". The results of the system application prove high practical value of the research and development activities carried out.
Azovtsev, D. Pokrovsky. ArgusSoft "Major principles of creating systems based on using precedents" /All-Russian Scientific and Practical Conference. - "Reengineering of Business Processes on the Basis of Modern Information Technologies" (April 23-24,1997). Collection of Scientific Studies. - Moscow 1997, pp. 79-85.
A. N. Gorban, D. A. Rossiev. Neuron Systems on PC. - Novosibirsk: Science. Siberian Publishing Company, RAS, 1996. - p. 276.