ALGORITHMS FOR CONSTRUCTION OF NEURAL NETWORK STRUCTURES FOR DETECTION AND RECOGNITION OF ATTACKS ON COMPUTER SYSTEMS

Branches of Industry : IT in different industries.

 

Brief description of the technique, technology and main features.   The goal of the researches is development of highly effective methods for detection and analysis of network intrusions on computer systems, capable to react on attacks in the real time and to find out unknown attacks. The scientific idea consists in generation and integration of various artificial neural networks in a single modular and/or multiagent intrusion detection system. Artificial neural networks have potential for solving different and difficult tasks from the domain of intrusion detection. Thus possessing the great flexibility and scalability capabilities, modular and multiagent neural networks become the effective tool in hands of a developer of computer security means.

 

Novelty of the product.  The technique for recognition of entrance patterns related to network activity is offered. The original neural network decisions for construction of the intrusion detection system are considered. The developed architectures of the intrusion detection system are architectures of higher level due to integration of separate neural networks in a single system.

 

Urgency. Existing approaches in the domain of intrusion detection do not guarantee an absolute protection, so the primary goal is an improvement of quality of detection and recognition process. Risks to lose confidential information due to unauthorized access are constantly increasing. So an active protection of computer networks in the real time will allow to provide information resources security.

 

Advantages in comparison with analogs. The developed methods show high efficiency in the domain of intrusion detection. A trained neural network functions quickly enough, what is especially important in the real time mode. Besides possessing high generalizing abilities neural networks can detect new attacks and effectively process noisy data. The results of our experiments correspond to a world level, and sometimes even exceed it due to the fact that modular and multiagent neural networks were applied. Such an approach is characterized by flexibility and scalability that is especially important for protection of computer systems.

 

Purpose. The results of the research can be used for protection of computer networks against attacks.

 

Field of application. Protection of information resources of an organization.

Контакты: 

 

Brest state technical university

For reception of the contact information on working out interesting you fill, please, the inquiry form.