IMMUNOLOGICAL MODEL OF KEYBOARD MONITORING OF INFORMATION SYSTEM OPERATORS

Abstract

The purpose of this work is to develop a model of keyboard monitoring of information system operators,
based on the use of a chain method of accounting keyboard handwriting parameters. The specified
method provides estimation of operator's keyboard handwriting on chains of characters of given
length, reflecting linguistically related parameters of keyboard set, characteristic for the given operator.
The keyboard typing of such chains by the operator with "good" keyboard handwriting has significantly
higher individuality due to correlation dependences between the time parameters of successive characters
and pauses. As a result, the chain method allows to provide higher accuracy of operator's identity
verification. Keyboard monitoring based on the chain method is proposed to be implemented in the
basis of artificial immune systems using an immunological model of clonal selection, in which the detectors
are represented by identifying parameters of the distribution area of the keyboard parameters of
"friend". In the tasks of keyboard monitoring the area of distribution of keyboard parameters of the
verified operator is always significantly less than the cumulative area of distribution of keyboard parameters
of other possible operators. The choice of the specified model allows to significantly reduce the
required volume of the detector population, and as a consequence - to significantly reduce the verification
time of the working operator. The decision to replace "friend" operator with "stranger" is proposed
to be considered reasonable when the frequency of operation of detectors exceeds the established
threshold value. The proposed immunological model has a number of advantages. The use of the chain
method of keyboard parameters accounting allows to verify the operator with greater accuracy in comparison
with traditional methods. The clonal selection model in combination with vector representation
of the keyboard data allows to significantly speed up the learning process and reduce the time required
to make a timely decision on the presence of a "stranger" operator. An important advantage of the model
is the ability to learn solely from the examples of keyboard handwriting operationally available
"friend" operators. The use of the clonal selection model also makes it possible to significantly reduce
the required volume of the population of detectors capable of effectively "covering" the distribution area
of the keyboard parameters of "friend" operator