Leon Reznik

GPU-based simulation of spiking neural networks with real-time performance & high accuracy

A novel GPU-based simulation of spiking neural networks is implemented as a hybrid system using Parker-Sochacki numerical integration method with adaptive order. Full single-precision floating-point accuracy for all model variables is achieved. The implementation is validated with exact matching of all neuron potential traces from GPU-based simulation versus those of a reference CPU-based simulation. A network of 4096 Izhikevich neurons simulated on an NVIDIA GTX260 device achieves real-time performance with a speedup of 9 compared to simulation executed on Opteron 285, 2.6-GHz device.

Anomaly Detection Based Intrusion Detection

This paper is devoted to the problem of neural networks as means of intrusion detection. We show that properly trained neural networks are capable of fast recognition and classification of different attacks. The advantage of the taken approach allows us to demonstrate the superiority of the neural networks over the systems that were created by the winner of the KDD Cups competition and later researchers due to their capability to recognize an attack, to differentiate one attack from another, i.e. classify attacks, and, the most important, to detect new attacks that were not included into the training set. The results obtained through simulations indicate that it is possible to recognize attacks that the intrusion detection system never faced before on an acceptably high level

ARTIFICIAL INTELLIGENCE APPROACHES FOR INTRUSION DETECTION

Recent research indicates a lot of attempts to create an intrusion detection system that is capable of learning and recognizing attacks it faces for the first time. Benchmark datasets were created by the MIT Lincoln Lab and by the International Knowledge Discovery and Data Mining group (KDD). A number of competitions were held and many systems developed as a result. The overall preference was given to expert systems that were based on decision making tree algorithms. This paper explores neural networks as means of intrusion detection. After multiple techniques and methodologies are investigated, we show that properly trained neural networks are capable of fast recognition and classification of different attacks at the level superior to previous approaches.

Neural networks for cognitive sensor networks

The paper puts forward a concept of cognitive sensor networks and investigates a feasibility of artificial neural networks application for its realization. It describes a design of novel hierarchical configurations imitating the structural topology of brain-like architectures. They are composed from artificial neural networks distributed over network platforms with limited resources. The paper examines a cognition idea based on its implementation through the signal change detection. The novel multilevel neural networks architectures are designed and tested in sensor networks built from Crossbow Inc. sensor kits. The results are compared against conventional multilayer perceptron structures in terms of both functional efficiency and resource consumption.

Automated Data Quality Assessment of Marine Sensors

The automated collection of data (e.g., through sensor networks) has led to a massive increase in the quantity of environmental and other data available. The sheer quantity of data and growing need for real-time ingestion of sensor data (e.g., alerts and forecasts from physical models) means that automated Quality Assurance/Quality Control (QA/QC) is necessary to ensure that the data collected is fit for purpose. Current automated QA/QC approaches provide assessments based upon hard classifications of the gathered data; often as a binary decision of good or bad data that fails to quantify our confidence in the data for use in different applications. We propose a novel framework for automated data quality assessments that uses Fuzzy Logic to provide a continuous scale of data quality. This continuous quality scale is then used to compute error bars upon the data, which quantify the data uncertainty and provide a more meaningful measure of the data’s fitness for purpose in a particular application compared with hard quality classifications. The design principles of the framework are presented and enable both data statistics and expert knowledge to be incorporated into the uncertainty assessment. We have implemented and tested the framework upon a real time platform of temperature and conductivity sensors that have been deployed to monitor the Derwent Estuary in Hobart, Australia. Results indicate that the error bars generated from the Fuzzy QA/QC implementation are in good agreement with the error bars manually encoded by a domain expert.

Intelligent Real-Time Adaptation for Power Efficiency in Sensor Networks

This paper presents an intelligent, dynamic power conservation scheme for sensor networks in which the sensor network operation is adaptive to both changes in the objects under measurement and the network itself. The conservation scheme switches sensor nodes between a sleep and an active mode in a manner such that the nodes can maximize the time they spend in a power-efficient sleep state, which corresponds to a nonmeasuring and/or nontransmitting mode, while not missing important events. A switching decision is made based on changes (or their absence) in the signals sensed from the environment by an intelligent agent that has been trained to determine whether or not a special event has occurred. This intelligent agent is based on a novel neural network topology that allows for a significant reduction in the resource consumption required for its training and operation without compromising its change detection performance. The scheme was implemented to control a sensor network built from a number of Telos rev. B motes currently available on the market. A few new utilities including an original neural network-based intelligent agent, a ldquovisualizer,rdquo a communication manager, and a scheduler have been designed, implemented, and tested. Power consumption measurements taken in a laboratory environment confirm that use of the designed system results in a significant extension of sensor network lifetime (versus ldquoalways onrdquo systems) from a few days to a few years.

Signal change detection in sensor networks with artificial neural network structure

The paper describes a design and implementation of a novel intelligent sensor network protocol enhancing reliability and security by detecting a change in sensor signals. The change could be caused by the sensor or communication unit malfunctioning or by malicious altering of a measurement result. The protocol utilizes a neural network function prediction methodology to predict sensor outputs in order to determine if the sensor outputs have changed. The parameter choice and the relationship between the threshold values and false alarm and missing attack rates are studied. The protocol is implemented and tested in real life environments with sensor networks built from Crossbow MICA motes. The test results are analyzed and recommendations on applications are provided

Fuzzy and probabilistic models of association information in sensor networks

The paper considers the problem of improving accuracy and reliability of measurement information acquired by sensor networks. It offers the way of integrating sensor measurement results with association information available or a priori derived at aggregating nodes. The models applied for describing both sensor results and association information are reviewed with consideration given to both neuro-fuzzy and probabilistic models and methods. The information sources, typically available in sensor systems, are classified according to the model (fuzzy or probabilistic), which seems more feasible to be applied. The integration problem is formalized as an optimization problem.

Notice of Violation of IEEE Publication Principles Which models should be applied to measure computer security and information assurance

The paper attempts to investigate a feasibility of developing new models and methodologies integrating probabilistic and soft computing techniques and applying them to measurement of computer system security. It reviews the methods, which have been developed and applied up to now for this purpose, and analyses their applicability. It concludes that neither of methodologies being applied for measurement of computer security and/or reliability may be considered as comprehensive and good. Summarizing the required features of the measurement models, the paper concludes that new synergetic models and approaches need to be developed.

Fuzzy models in evaluation of information uncertainty in engineering and technology applications

The paper studies the problem of information uncertainty evaluation in modern engineering and technology applications, and especially the system design. It analyses the virtual environment design and engineering measurement. Information typical for those applications is classified according to its uncertainty types. Uncertainty sources are identified. Fuzzy theory models are proposed. Examples of their applications in characteristic problems are given.