Modeling Fault Tolerant and Secure Mobile Agent Execution in Distributed Systems - Distributed Artificial Intelligence, Agent Technology, and Collaborative Applications - page 144

conclusion
In this chapter, we have presented the mobile code paradigm, which is a collection of remote evalua-
tion, code on demand, and mobile agents, as an alternative to the conventional client/server paradigm.
We examine security concerns of the mobile code paradigm, and survey existing security attacks and
mechanisms to evaluate the current status of mobile code security. We conclude that the mobile code
paradigm is still to be developed with respect to its security aspects and that mobile agent protection
needs particular attention. To investigate the security threats to mobile agents, we implemented a simple
Traveling Information Agent System, and discussed the possible attacks to the agents in this system,
based on the attack model in [26].
We have identified two important properties for fault-tolerant mobile agent execution: non-blocking
and exactly-once. Non-blocking ensures that the agent execution proceeds despite a single failure of
either agent, place, or machine. Blocking is prevented by the use of replication. This chapter discussed a
mobile agent model for processing transactions, which manipulate object servers. An agent first moves
to an object server and then manipulates objects.
General possibilities for achieving fault tolerance in such cases were discussed and the respective
advantages and disadvantages for mobile agent environments and the intended parallel and distributed
application scenarios were shown. This leads to an approach based on warm standby and receiver side
message logging. We have used dynamically changing agent domains to provide flexible, adaptive and
robust operation. The performance measurement of Fault-Tolerant Mobile Agent System shows the over-
head introduced by the replication mechanisms with respect to a non-replicated agent. Not surprisingly,
it also shows that this overhead increases with the number of stages and the size of the agent.
r eferences
Aguilera, M. K., Chen, W. & Toueg, S. (2000). Failure detection and consensus in the crash-recovery
model. Distributed Computing, 13 (2), 99-125.
Brocklehurst, S., Littlewood, B., Olovsson, T., & Jonsson, E. (1994). On measurement of operational
security. In Proceedings of the Ninth Conference on Computer Assurance (COMPASS'94): Safety, Reli-
ability, Fault Tolerance and Real Time, Security (pp. 257-266).
Chan, H. W., Wong , K. M., & Lyu, R. (1993). Design ,implementation ,and experimentation on mobile
agent security for electronic commerce application. In S. Mullender (Ed.), Distributed systems (2nd ed.)
(pp. 199-216), Reading, MA: Addison-Wesley.
Chess, D., Harrison, C. G., & Kershenbaum, A. (1998). Mobile agents: Are they a good idea? In G. Vigna
(Ed.), Mobile agents and security (pp. 25-47). Springer-Verlag.
Defago, X., Schiper, A. & Sergent, N. (1998, October). Semi-passive replication. In Proceedings of the
17th IEEE Symposium on Reliable Distributed System (SRDS'98) (pp. 43-50).
Fischer, M. J., Lynch, N. A. & Paterson, M. S. (1983, March). Impossibility of distributed consensus
with one faulty process. In Proceedings of the second ACM SIGACT-SIGMOD Symposium: Principles
of Database System (p. 17).