2006 Fiscal Year Final Research Report Summary
Immunity-Based Complex Systems and Its Applications to Diagnosis and Repair by Sensor Agents
| Project/Area Number |
16300067
|
|---|
| Research Category |
Grant-in-Aid for Scientific Research (B)
|
| Allocation Type | Single-year Grants |
|---|
| Section | 一般 |
|---|
| Research Field |
Sensitivity informatics/Soft computing
|
|---|
| Research Institution | Toyohashi University of Technology |
|---|
Principal Investigator |
ISHIDA Yoshiteru Toyohashi University of Technology, Department of Knowledge-Based Information Engineering, Professor, 工学部, 教授 (80159748)
|
|---|
| Co-Investigator(Kenkyū-buntansha) |
WATANABE Yuji Nagoya City University, Graduate School of Natural Sciences, Lecturer, 大学院システム自然科学研究科, 講師 (60314100)
|
|---|
| Project Period (FY) |
2004 – 2006
|
| Keywords | Immune Systems / Complex Systems / Sensor Agents / Self-Information Acquisition / Repair / Probabilistic Cellular Automata / Spatial Prisoner's Dilemma / Immunity-Based Systems |
|---|
| Research Abstract |
This project built a sensor agent system mimicking the immune system. With the system, the following three results have been obtained.
1.The'self-information'acquisition: Sensor agents mounting specific recognition capability of the target units are introduced. Sensor agents monitor not only the target state but also the other agents that monitor other states, which have a certain relation with the original target state. The relation (or constraint) when the target system is normal is considered as the `self-information'; and the relation when the system is abnormal as the nonself-information. By applying the technique to a certain domain of diagnosis, we have shown that the real-time diagnosis can be done using only the 'self-information'that has been acquired off-line beforehand. However, it has been also revealed that the 'self-information'undergoes change depending on the environment. The role and the impact of the diversity of agents have been also studied by a multi-agent simulati
… More
on. Further, it has been shown that the sensor agents can have a collective recognition capability, which includes a computational capability that can solve some combinatorial problems.
2.The self-repairing system: The network cleaning problem is formalized with a model of agents mounting not only recognition capability but repairing by overwriting the state of the target node with the self content. The model has been shown to include a certain class of the probabilistic cellular automaton in a percolation theory. Critical phenomenon has been also observed with the parameter of success rate in overwriting. The impact of recognition capability has also been studied with a regular lattice network. Repairing strategies are devised by spatial strategies in the Spatial Prisoner's Dilemma. The repairing can be controlled by spatial strategies when the network is inhomogeneous in a spatial as well as in a temporal sense.
3.Applications of the system: A system for designing, developing and implementing sensor networks involving sensor agents has been developed. The system allows deploying the sensor agents in networks when they can be simulated in the virtual network environment. With a framework of eliminating the nonself by dynamically identifying self and nonself, mathematical models such as a hidden Markov model has been naturally introduced in the network security applications such as intruder detection systems and masqueraders detection systems. Less
|
