Adaptive Touch Interface: Application for Mobile Internet Security

Abstract

In modern means of mobile Internet security, including those based on touch screens, various visualization models are used. However, with the increasing complexity of these models, the requirements for models of user interaction with visualization change, the need for their adaptability increases. The article proposes an adaptive approach to the formation of a user interface based on touch screens for managing mobile Internet security. The results of experiments on user interaction with visualization of a centralized and decentralized network of devices and user perception of certain gestures when using touch screens are also shown. The problems and advantages of this type of interface, identified during the tests are described.

Full text

Adaptive Touch Interface: Application
for Mobile Internet Security
Ksenia Zhernova , Maxim Kolomeets(B
), Igor Kotenko ,
and Andrey Chechulin
St. Petersburg Federal Research Center of the Russian Academy of Sciences
(SPC RAS), St. Petersburg, Russia
{zhernova,kolomeec,ivkote,chechulin}@comsec.spb.ru
Abstract. In modern means of mobile Internet security, including those
based on touch screens, various visualization models are used. However,
with the increasing complexity of these models, the requirements for
models of user interaction with visualization change, the need for their
adaptability increases. The article proposes an adaptive approach to the
formation of a user interface based on touch screens for managing mobile
Internet security. The results of experiments on user interaction with
visualization of a centralized and decentralized network of devices and
user perception of certain gestures when using touch screens are also
shown. The problems and advantages of this type of interface, identified
during the tests are described.
Keywords: Graphical user interface ·Information security ·Touch
interface ·Adaptive interfaces ·Predictive interfaces ·Touch screen
1 Introduction
One way to analyze security is through visual analytics. Visual analytics uses
data visualization to detect events, interpret incidents and select countermea-
sures. In mobile internet security, data visualization has many uses: access con-
trol in various security models; analysis of the state of the networks formed by
Internet of Things (IoT) devices; analysis of mobile security metrics and others.
To solve the problems of this kind, specialists use various visualization models
that contain traditional interaction interfaces. Nevertheless, the complication
of visualization models requires new forms of interaction that would be more
convenient for the operator and thereby increase the speed and quality of decision
making. One such solution is touch screen interfaces. However, they are usually
not considered as a tool for the interaction of the analyst and data visualization
mechanisms.
Another problem is the contradiction between the functionality implemented
in the interface and the functionality necessary for the user to solve specific
Supported by the grant of RFBR 18-07-01488-a in SPIIRAS.
c
Springer Nature Singapore Pte Ltd. 2020
I. You et al. (Eds.): MobiSec 2019, CCIS 1121, pp. 53–72, 2020.
https://doi.org/10.1007/978-981-15-9609-4_5

54 K. Zhernova et al.
tasks. To solve this problem, adaptive and predictive interfaces are used. They
adapt to a specific user and the task that he/she is solving.
In this paper, we propose the approach to the formation of user interfaces
based on touch screens and recognition of operator gestures. The advantage of
this approach (in comparison with the traditional interface) is that it will increase
the speed of mobile internet security events management, simplify interaction
with visualization models, and improve the quality of decision making. To do
this, we provide models of user-interface interaction and interface adaptation
algorithms for the following tasks of mobile internet security: managing a hier-
archical centralized network of embedded devices and visualizing a decentralized
sensor network.
The scientific novelty of this paper is the proposed combined approach for
implementing a touch interface based on: (1) adaptive adjustment algorithm
for a specific user and the mobile internet security task, (2) the use of “best
practices” to form a predictive gesture interface. The contribution of this paper
is the approach that includes models and algorithms for adapting the touch
interface to the tasks of mobile internet security.
This paper has the following structure. The second section provides the
overview of related works on the field of visual analytics of mobile internet secu-
rity and interaction interfaces. The third section presents the approach to the
development of the adaptive interface for mobile internet security. In this section
the business model and the practical model of user interaction and visualization,
the adaptation algorithm and the algorithm of “best practices” gesture design are
proposed. The fourth section describes experiments on the perception of gestures
by users using examples of visual analytics for the tasks of managing a hierar-
chical centralized network of embedded devices and visualizing a decentralized
sensor network. The fifth section considers the advantages and disadvantages of
the proposed approach and describes the further direction of work.
2 State of the Art: Usability Approaches
Human-computer interaction interfaces are closely related to visualization mod-
els. So, depending on the task, various types of visualization models are used.
The implementation of the interfaces depends on visualization models. For exam-
ple, using graphs, one can visualize a computer network [22], port scanning [7],
attacks and their routes [9,20,21], and it is also possible to simulate attack sce-
narios [13]. At the same time, visualization methods can be combined with each
other. For example: using a graph of the “tree” type one can depict the physical
hierarchical topology of a computer network; a radial tree can be used to visu-
alize attacks; Chord diagrams – to simultaneously display physical and logical
topologies; and matrices – to display the availability of network segments for an
attacker [15].
For access control, such visualization methods can be used for visualization
of relationship between subject and object in a specific access right model. So,
for discretionary access models, matrices are used [11]. For Take-Grant access

Adaptive Touch Interface: Application for Mobile Internet Security 55
models [8], graphs are used. For hierarchical RBAC models, TreeMaps are used
[14]. In addition, there are complex visualization models that are designed for
analysis in combined security models. For example, triangular matrices [18]use
visualization of both – matrices and trees.
Each of the existing models is used in a specific case of analysis and manage-
ment of access rights. The more complex the visualization model and the more
complex the analysis of the security model, the more complex the interaction
methods the operator needs. For example, in [11] the access matrix is presented,
it uses the mechanisms of filtering and grouping of subjects and objects. To do
that matrix uses classic tools, for example, drop-down lists. In TreeMaps, one
can filter data by showing only a specific part of the tree [14]. To do this one
need to click on the root of the specific subtree.
When analyzing the state of networks the analysts use graphs, TreeMaps,
matrices, and other visualization models [15]. Graphs are the most universal,
and with their help one can visualize any network structure [18]. TreeMaps are
suitable for visualizing hierarchical networks [9]. Matrices are used for almost
fully connected networks [22]. Also for networks that can form planar structures,
Voronoi maps are applied [19]. An example of such a network can be a self-
organizing sensor network, the topology of which was reduced to planar in order
to save energy and reduce interference [19]. Each method has its own advantages
and disadvantages; therefore, they can be used together [23].
The presented visualization models are realized in many fields. As already
mentioned, the more complex the task and the more metrics are needed to visu-
alize, the more complex the visualization model becomes. For example, [17]and
[16] presented approaches to combining visualization models in order to display
more metrics. On the other hand, the more complex the model, the more inter-
action tools an operator needs. For example, when implementing 3D models, one
should include tools that implement rotation and scaling. In overloaded graphs,
it is also suitable to implement scaling tools [17]. Moreover, often standard tools
may not be enough, and instead of standard scaling, fisheye [24] and Cartesian
distortion can be used. All this leads to an overloaded interface and complicates
the work of the operator-analyst.
When designing visual models, only traditional control methods based on
the use of a monitor, a mouse, and a keyboard are considered. However, visual
analytics can also be carried out using tablets, smartphones and other devices
with touch screens, as they are becoming more widespread and provide greater
operator mobility, for example, in production. The papers devoted to visual
analysis of information security do not discuss approaches based on touch screens
and how they can affect the process of visual analysis and decision-making of
information security.
Security analysis applications that have a touch interface implementation are
not common. We examined those few of them (for example, “Network Scanner”,
“Net Analyzer” and “IP Tools”) and found that gestures are most often limited to
touching one finger (rarely two), while the interaction with visualization models is
also limited to clicking and dragging. Thus, the interfaces of many modern appli-
cations serve only for a simple imitation of interaction with a computer mouse.

56 K. Zhernova et al.
For information security applications, due to the complexity of the processed
information and the complexity and multi-level visualization of data, standard
gestures that mimic a mouse and keyboard may not be enough. However, gestures
should also not be too complicated to remember or unnatural to use. In order to
use such gestures in security interfaces, we propose the approach that is based
on models of user interaction and visualization, an adaptation algorithm, and a
best practices gesture design algorithm.
3 Adaptation Technique
In order to understand how the interface of mobile internet security applications
can work, one should pay attention to the specifics of information security inter-
faces, how they differ from others. In mobile internet security applications, and
in general information security, the following elements are common:
– using the color of current events in three colors to distinguish between the
degree of danger – green (safe, for example, the embedded device is charged),
yellow (medium danger, for example, the charge level of the embedded device
is coming to an end), red (the highest degree of danger, for example, the
embedded device is turned off);
– nesting (request of additional parts on demand, for example – showing device
parameters on a graph);
– a large amount of data that needs to be processed (for example, traffic routes);
– situational awareness (providing the user with relevant data with reference
to time and place, for example, when monitoring the network online);
– visualization of the processed data (for example, the presentation of various
network topologies).
The listed elements must be present in the applications of mobile internet secu-
rity. However, their visual presentation and interaction methods may vary. The
example is as follows: with frequent reports of security risks marked in red, the
user may get tired and begin to ignore them. This problem can be solved using
the adaptability of the interface – at certain intervals to change the tone of the
alarm message within the red color, for example, use the shade of the red color
“magenta”. The user will notice the changes and will begin to pay attention to
the messages again. Thus, adapting an interface is also a necessary part of its
design.
Adaptive interface is an interface that adapts to the needs of the user based
on his/her behavior when working with the application. The adaptive interface
often refers to an adaptive design that is modified according to the resolution of
the user’s device, for which the flexible grid-based templates are used (a set of
open Bootstrap libraries can be an example of this).
In addition, there are so called predictive interfaces. A predictive interface
is able to predict what action the user is about to take at the next moment,
as well as which interface design will be most convenient for the user based on

Adaptive Touch Interface: Application for Mobile Internet Security 57
his/her behavior. The implementation of the predictive interface is possible, for
example, on the basis of neural networks or the collection of statistics on user
actions. A simple example of a predictive interface is predictive typing, in the
presence of which the system remembers words and word combinations most
often used by the user.
This section proposes an approach to designing an adaptive application inter-
face of mobile internet security, which allows the users to adjust the system of
interaction with the application for themselves and minimize the need to adapt
themselves.
In order to understand what place the interfaces occupy in the process of
visual analysis, it is necessary to determine the model of interaction between the
user and the visualization module. At the level of business logic, the model is as
follows (Fig. 1).
Fig. 1. Business logic of the interface model.
This model assumes that the user will interact with the visualization module
through gestures, the system will process user commands that are implemented
through gestures. Moreover, each specific user has its own characteristics, which
the system will also process, therefore, as a result, not only a visual representa-
tion will be formed, but the result of adaptation for a specific user.
At the implementation level, the model looks as shown in Fig. 2.
Data comes from a computer system and is downloaded to the application,
processed, displayed and drawn to get the final visualization. At the same time,
the user can interact with the image by gestures. And while user interacts with
images, the system processes, performs adaptation to the particular user’s fea-
tures. So, the image can be rendered again and then it will be modified to adapt
it for the specific user.
As one can see, the key elements of interaction are the processes of informa-
tion output and input, which are carried out using visualization and gestures,
respectively. To adapt them, interaction processes should be considered at two
levels: (1) at the level of interaction between a machine and a person; (2) at the
level of interaction between a person and a machine;

58 K. Zhernova et al.
Fig. 2. A visualization model that includes a gesture interface.
For this, we propose the adaptation algorithm, the idea of which is that the
visualization model should independently recognize the combination of gestures
and functions that are most convenient for the user. The interface adaptation
algorithm can be divided into two following stages.
The first stage consists of the following steps.
1. Adjustment in the initialization process for a person or group. The initial-
ization phase usually involves the user entering the system, determining the
level of preparation of the user, as well as the issuance of current information.
The following rules can be mentioned as the “best practices” from this point
of view [4]:
–focusing the user’s attention on where to start. This item includes an idea
of what elements need to be made larger, highlighting headers, etc.;
–visual hierarchy of interface elements involves the use of one column where
possible, avoiding unjustified voids inside any interface element, partially
overlapping some design elements with others to achieve the integrity of
the perceived material;
–the correct grouping of elements for reasons of similarity, proximity, clo-
sure, connections between them and continuity. An example is the group-
ing of similar functions, the separation of functional elements from each
other by space, etc.;
–displaying changes so that it is noticeable to the user (animation mini-
mizing the window, displaying an incorrectly filled field in red, etc.);
–refusal of unnecessary information – to provide the user with the oppor-
tunity to hide information that he/she rarely uses, remove self-evident
instructions, inscriptions and uninformative pop-ups, hide functions that
are rarely used;

Adaptive Touch Interface: Application for Mobile Internet Security 59
Fig. 3. Examples of diagrams created using D3.js that can be used for applications of
mobile internet security.
–“approaching” important functions and frequently used data to the user
(the default settings should be the most frequently used, the most frequent
answers in this segment are displayed at the top of the drop-down list,
etc.);
–providing important information and prompts on demand (for example,
when one hover or hold one’s finger on the touch screen);
–visualization of primary information on the home page;
–display of many ways to accomplish a task (for example, provide the
ability to enter the system by mail, using a login or phone number);
–providing a hint of the required actions, including the designation of
mandatory and optional actions, about how the result of the actions
should look;
–emphasizing of elements which the user can interact with (by highlighting,
adding icons with actions, etc.);
–help users to avoid errors (for example, by displaying available options,
structuring text fields, easy ways to exit the option, informative error
messages).

60 K. Zhernova et al.
2. Global adjustment. When a user interacts with an interface, a number of prob-
lems may arise due to sociocultural differences. For example, the perception
of the semantics of color may be different in different cultural environments
– different symbols and pictograms can be used for the same purposes. So, in
most European countries it is customary to mark the correct answer with a
check mark, usually green, but in some countries (particularly in Japan and
South Korea) the red circle indicates the correct answer. Thus, an example of
global adjustment, in addition to changing the application language, can be a
change in the color scheme of the interface, the arrangement of windows, etc.
A similar global adjustment is also necessary when it comes to any physical
difficulties, for example, disorders of color perception [6].
3. Adjustment for a post. Obviously, all employees of an organization cannot
be experts in information security. For this reason, it is advisable to allow
the user to choose the type of interface in accordance with his/her position
in order to see the details that are necessary for him/her. For example, a
developer may request more details related to program code. An information
security specialist may not have developing skills. However, security details
will be important to him.
4. Adjustment for a specific person. This type of adjustment lies in the individ-
ual way of the user to work with the application. An individual manner can
be expressed in a specific choice of the most convenient visualization models
for the user, individual perception of gestures of the touch interface, the way
to perceive and analyze information from the system. Also, this may include
the most frequently used application functionality and the most frequently
requested information.
The second stage of adaptation occurs in the process. It can be divided into
the following two components.
1. Adjustment for interaction from the computer. This component relates to
the visualization of information processed by the program. The adjustment
may be the selection of the most comfortable visual models for the user, the
adjustment of the selected color scheme, the selection and tuning of signals
other than visual ones – sound and vibration. If necessary, more detailed
information is displayed, details on demand, prompts, etc. Figure3shows a
set of complex visualization models built on the basis of the D3.js library (for
the Javascript language) which can be used to solve mobile internet security
tasks.
2. Adjustment for human interaction. This component relates to how a person
communicates his/her intentions to a software and hardware system. The
proposed approach assumes that at this stage the system should determine
which gestures of the touch screen it is more convenient for a person to use
for certain functions, adapt to the execution of the gesture of a particular
person (it may differ for different users: for example, when using multi-touch
screens they do not put a few fingers on screen at the same time, the pressing
time is different, the user can start making a one gesture, then change his/her
mind and finish with another).

Adaptive Touch Interface: Application for Mobile Internet Security 61
A feature of touch interfaces, in turn, is interaction through gestures. The gesture
interface, like the graphic one, should follow the principle of direct manipulation
[12], i.e. used gestures should be intuitive to the user. An example of the use of
intuitive gestures is shown in Fig. 4. The user can also reconfigure gestures at its
discretion.
Fig. 4. Examples of simple and complex gestures for touch interfaces presented in [1].
Also, to improve gestures, we provide the algorithm that allows one to create
gestures in accordance with their “best practices”:
(1) adaptation to a mobile device (changing the page width, text and picture
size when changing the screen resolution, the possibility of scrolling to the
side or scrolling down) [3];
(2) creating graphic elements in such a way that it is convenient to interact with
gestures (large enough buttons, paging elements, high-resolution images so
that they can be enlarged, the absence of a large number of small elements
in a row that one needs to press) [3];
(3) using standard gestures, such as tapping, double tapping, dragging, scrolling,
swiping in any direction, pinching in and out by two fingers, pressing, twist,
rotating or shaking the device [2];
(4) using gestures that are intuitively appropriate for each function [2];
(5) rejection of the traditional computer mouse hover and gestures associated
with the mouse when developing a gesture version of the interface [5];
(6) creating interface elements that will not be overlapped with the hands of
the user, the user should be able to see these elements [5].

62 K. Zhernova et al.
4 Implementation
The proposed models of human-computer interaction were implemented as a soft-
ware prototype of a web application. The prototype was executed in JavaScript
using the HTML5 markup language and the D3.js, hammer.js libraries and the
free Bootstrap package. The project consists of two components: a visualization
component and a component of human-computer interaction.
Four tests of two types were carried out. Two tests of the first type (Test A)
were based on user interaction with the touch screen. Two tests of the second
type (Test B) were based on interaction with the traditional hardware interface
(keyboard and mouse). Test 1 and Test 2 were formed on the basis of two different
datasets (Data Set 1 and Data Set 2, respectively) and differ in graphs.
The first test contained an image of a decentralized sensor network graph,
the second test contained a hierarchical centralized network graph. Thus, the
following tests were carried out:
1) Test 1 A - decentralized graph and touch screen; 2) Test 2 A - centralized
graph and touch screen; 3) Test 1 B - decentralized graph and keyboard and
mouse; 4) Test 2 B is a centralized graph and keyboard and mouse.
Data Set 1 (Test 1) – Visualization to simulate a decentralized sensor network
without reducing to planarity. The experiment used the data on the simulation of
a decentralized sensor network, which consists of autonomous devices. As part of
the simulation, the following device parameters were taken into account: battery
charge, light and sound levels. Some of them were outdoors. Each device has a
critical level, which was calculated based on the criticality of the assets [17] that
were located in this area. Thus, a loss of a sensor would mean a loss of control
over this asset. Since the devices are autonomous, they are discharged, but they
can be charged using solar panels.
Data Set 2 (Test 2) – Visualization to simulate an integrated security system
hierarchical centralized network containing embedded devices [10]. Embedded
devices are equipped with a set of sensors: motion sensors, RFID reader, com-
bustible gas sensor, window breaking sensor, temperature, humidity and light
sensor. Embedded devices were connected to a hub, which collected, normalized,
and pre-processed the received data. Hubs connected to a server whose task is
to store, process, analyze security messages from devices and the status of these
devices.
When using visualization in the analysis process, it is possible to interact
with visualized information through gestures on the touch screen. The following
gestures were implemented in the prototype:
– attracting the nearest vertex of the graph (device) and calling the context
menu for this vertex when touching with a finger, selecting the context menu
option by repeated touch. Selecting individual vertices and vertex groups is
implemented through the context menu;
– moving three fingers left/right – calling/hiding additional information
(show/hide MAC addresses, charge level, number of transmitted messages,
etc.);

Adaptive Touch Interface: Application for Mobile Internet Security 63
– four-finger touch – filtering change (display the vertex color as the type
of device, device charge level, number of transmitted messages, number of
received messages, etc.);
– pinching in and out of five fingers – changing graph connections (show how
devices are physically connected, as well as show their traffic routes).
Gestures were originally assigned to certain functions that the application per-
forms. On the application page on the right side is the explanation of the corre-
spondence of gestures to functions.
As a test of this prototype, a number of tasks were proposed based on the
available methods of human-computer interaction.
For data set 1:
1. Attach devices with specific MAC addresses (MAC addresses are hidden and
shown by a specific gesture).
2. Highlight discharged devices (high charge, device almost discharged, device
discharged and turned off – are set by color).
3. Highlight a specific type of device (the type of device is set by color).
4. Highlight devices that are not connected to a self-organizing network (a vertex
without edges).
5. Highlight almost discharged devices with high criticality (the device is almost
discharged is set by color, the criticality of the asset is determined by the
vertex size).
For data set 2:
1. Attach all switched off devices (switched on devices, switched off – are set by
color).
2. Highlight all hubs (the type of devise is set by color).
3. Highlight all RFID scanners and smoke detectors based on the color of the
vertex (the type of detector is set by color).
4. Highlight the hubs that received the most messages (the more messages, the
larger the vertex).
5. Highlight the devices that generated the largest number of messages (the
more messages, the larger the vertex).
The prototype was launched through a browser on a PC with a touch screen
(Fig. 5and Fig. 6).
In Fig. 5, one can see the force graph, which is the simulation of a decen-
tralized sensor network without reducing to planarity. The vertices of the graph
are autonomous devices. The color indicates the type of device. The network is
self-organizing, interconnected devices have links, unconnected ones do not have
links.
Figure 6is the force graph denoting a simulation of an integrated security
system of hierarchical centralized network. The vertices of the graph are detectors
(yellow and white), embedded devices (purple), hubs (green), server (blue).

64 K. Zhernova et al.
Fig. 5. Appearance of the implemented web application, Data Set 1. Unconnected
devices have no links; type of device is set by color.
Fig. 6. Appearance of the implemented web application, Data Set 2. The colors of
vertices are detectors (yellow and white), embedded devices (purple), hubs (green),
server (blue). (Color figure online)
At the same time, control was carried out through this display through ges-
tures. The verification was carried out as follows:
1. The subject approached the stand.
2. The subject was instructed to interact with the task management interface,
which took 3 min to read.
3. The subject was explained how to go from test 1 (decentralized graph) to test
2 (centralized graph) and how to adjust the height and tilt of the screen.
4. Then the subject sat down to perform one of the tests.
At the same time, tests with a touch interface and tests with a traditional
interface were passed different people. The same person was forbidden to pass

Adaptive Touch Interface: Application for Mobile Internet Security 65
both types of tests. The test observer was responsible for the equipment and
fixed the problems associated with it. At the same time, the observer was
forbidden to answer questions regarding the specifics of the test (interaction
with the visualization model itself).
5. During the execution of the next task it was required to interact with the
visualization.
6. After completing the last task, the “Finish” button should be touched; this
action initiated the download of a text file with the task execution logs to the
computer.
7. Then the collected logs were analyzed for the time spent on each task, as well
as the quality of the tasks (correct execution).
The tasks are divided into three groups (selection, interaction with the menu
and action) and their combinations (selection + menu, selection + action). Issues
related to the selection suggested the possibility of selecting one or more visual-
ization elements, interaction with the menu implied interaction with the drop-
down list options, the action was carried out using more complex gestures.
As a result of the experiment, the distributions of speed of the tasks were
obtained (Fig. 7and Fig. 8). The results were evaluated according to three
parameters: the maximum of distribution, the upper quantile (75% of the best
indicators), and the average value. For this, the distribution graphs were visual-
ized in the form of box-plot.
Fig. 7. Test 1: task execution speed in seconds for a decentralized graph, where a–
touch screens, b–traditional interface.

66 K. Zhernova et al.
Fig. 8. Test 2: task execution speed in seconds for a centralized graph, where a–touch
screens, b–traditional interface.
The speed of answering questions was determined as the difference between
the beginning of the answer (when the task text appeared in the corresponding
window) and the end of the answer (when the user clicked on the button for
moving to the next question). The time on the chart is measured in seconds.
Each chart compares the performance of tests with a touch interface (question
numbers with the letter “a”: 1a, 2a, etc.) and tests with a traditional button
interface (question numbers with the letter “b”: 1b, 2b, etc.). The same questions
are shown in the same color.
Tasks 1, 2, 3, 4, 9, 11 and 12 were devoted to the interaction with individual
vertices of the graph, tasks 5 and 6 involved interaction with a group of ver-
tices and tasks 10 and 13 contained interaction with additional information on
demand. All tasks except 7 and 8 had similar objectives and tasks 7 and 8 tasks
were different for centralized and decentralized graphs. In the case of a decen-
tralized graph, it was required to interact with the connections of the graph:
change the connections of the graph (task 7), and then fix several vertices and
change the connections again (task 8). In the case of a centralized graph, it was
required to interact with a group (task 7) and with separated vertices (task 8).
Below are tables (Table 1and Table 2) comparing the parameters of time
distributions when performing test tasks. The maximum distribution (Upper
fence), the upper quantile (Q3) and the average value (mean) are compared.
The comparison was carried out according to the following principles:

Adaptive Touch Interface: Application for Mobile Internet Security 67
Table 1. The efficiency of task groups in tests 1 (decentralized network)
Test 1 A Test 1 B
Tas k Parameter Val u e Parameter Va l u e
Upper fence 54 Upp er fence 147
Q3 25 Q3 73
1
Mean 22.68 Mean 58.5
Upper fence 36 Upp er fence 74
Q3 22 Q3 59
2
Mean 21.68 Mean 33.83
Upper fence 34 Upp er fence 44
Q3 19 Q3 22
3
Mean 15.07 Mean 16.44
Upper fence 22 Upp er fence 44
Q3 14.5 Q3 30
4
Mean 18.82 Mean 21
Upper fence 30 Upp er fence 50
Q3 22 Q3 39
5
Mean 15.5 Mean 35.5
Upper fence 35 Upp er fence 30
Q3 23 Q3 28
6
Mean 20 Mean 24.17
Upper fence 47 Upp er fence 28
Q3 28.5 Q3 22
7
Mean 24.29 Mean 19.56
Upper fence 90 Upp er fence 139
Q3 48 Q3 75
8
Mean 46.29 Mean 63.33
Upper fence 41 Upp er fence 72
Q3 29.5 Q3 53
9
Mean 23.11 Mean 41.78
Upper fence 45 Upp er fence 141
Q3 23 Q3 78
10
Mean 15.21 Mean 57.5
Upper fence 25 Upp er fence 33
Q3 14 Q3 22
11
Mean 16 Mean 16.39
Upper fence 18 Upp er fence 29
Q3 12.5 Q3 18
12
Mean 9.75 Mean 15.78
Upper fence 21 Upp er fence 14
Q3 11 Q3 12
13
Mean 10.48 Mean 12.22

68 K. Zhernova et al.
Table 2. The efficiency of task groups in tests 2 (centralized network)
Test 2 A Test 2 B
Tas k Parameter Val u e Parameter Va l u e
Upper fence 34 Upp er fence 24
Q3 26 Q3 18
1
Mean 21.68 Mean 16.39
Upper fence 42 Upp er fence 15
Q3 21.5 Q3 11
2
Mean 17.07 Mean 9.56
Upper fence 19 Upp er fence 23
Q3 11 Q3 15
3
Mean 13.42 Mean 10.78
Upper fence 23 Upp er fence 14
Q3 12.5 Q3 12
4
Mean 9.93 Mean 9.89
Upper fence 13 Upp er fence 11
Q3 9Q3 8
5
Mean 7.75 Mean 7.06
Upper fence 26 Upp er fence 14
Q3 15.5 Q3 10
6
Mean 12.25 Mean 9.67
Upper fence 19 Upp er fence 18
Q3 13 Q3 13
7
Mean 15.14 Mean 10.83
Upper fence 31 Upp er fence 29
Q3 20 Q3 19
8
Mean 20.43 Mean 17.06
Upper fence 22 Upp er fence 21
Q3 17 Q3 16
9
Mean 14.29 Mean 15.44
Upper fence 19 Upp er fence 11
Q3 12 Q3 10
10
Mean 9.04 Mean 8.5
Upper fence 13 Upp er fence 20
Q3 10 Q3 13
11
Mean 9.11 Mean 11.11
Upper fence 23 Upp er fence 17
Q3 14 Q3 13
12
Mean 13 Mean 12.11
Upper fence 13 Upp er fence 9
Q3 8Q3 7
13
Mean 6.11 Mean 6.89

Adaptive Touch Interface: Application for Mobile Internet Security 69
1. If the time difference is more than 3 s in favor of touch screens (the task exe-
cution time on the touch screen is shorter than on the traditional interface),
the results are considered good and cells are highlighted in blue.
2. If the time difference is less than 3 s in favor of either of the two tests, the
results are considered the same and cells are highlighted in yellow.
3. If the time difference is the traditional interface (the task execution time on
the touch screen is longer than on the traditional interface), the results are
considered unsatisfactory and cells are highlighted in red.
The final result is marked in a predetermined color in the cell with the task
number according to the principle of the majority element:
1. If most of the parameters are the same, the overall result is considered accept-
able and is marked in yellow.
2. If most of the parameters are “good”, the overall result is considered good
and marked in blue.
3. If there is at least one “unsatisfactory” parameter, the results cannot be
considered good and are considered: (1) acceptable if the “unsatisfactory”
parameter is one, (2) “unsatisfactory” if there are two or three “unsatisfac-
tory” parameters.
For a decentralized network graph, the touch interface showed the best result
in almost all test categories. The exception was task number 7, “Change com-
munications with the graph,” which was carried out by mixing/raising several
fingers across the screen. Otherwise, the results are better, or comparable to the
traditional interface. The experiment showed that the presented approach will
allow faster analytics of the self-organizing sensor network. Gestures allow one
to quickly and intuitively switch between metrics, capture the interested vertices
or groups of vertices, and switch between graph representations. Thus, the qual-
ity is improved, and the speed of decision-making in the management of mobile
networks is increased. The intuitive nature of gestures allows one to remember
more commands. It gives the possibility to analyze a larger number of metrics,
as it becomes easier to switch between them. Thus, in the process of managing
the mobile network, more useful information for decision making can be used.
For a centralized network graph, the touch interface showed a predominantly
equal result. The exception was task No. 1, “Pull and fix any point,” No. 2,
“Increase the selected point,” and No. 6, “Select all green points.” Otherwise,
the results are considered the same.
It is supposed to further implement an adaptive interface based on the collec-
tion of statistics on user actions, for example, what gestures for which functions
he/she uses most often.
5 Discussion
This research focuses on touch screen gestures as a way to improve the interaction
between the user and information security systems. Further work involves the

70 K. Zhernova et al.
development of a technique for creating adaptive cybersecurity interfaces for
touch screens.
The advantages of this approach are the following points:
1. All previous settings for a specific user are saved.
2. The interface configured for a particular user becomes more convenient for
that user, therefore, working with the software also occurs more quickly, effi-
ciently, with fewer errors.
3. Creating an adaptive gesture interface will allow the user to bind certain ges-
tures that are convenient for him/her to the existing functions of the appli-
cation.
4. Such an improvement will increase the speed of learning the interface of the
new application, as well as increase the speed and efficiency of the operator’s
further work with the information security application.
Possible disadvantages of the approach may include the following circum-
stances:
1. It will be difficult for another user to start work on the same device. In the
case of several people working in turn for one device, this approach will be
more likely a disadvantage. However, most often each employee has his/her
own individual workplace.
2. It will take some time until the application collects the necessary statistics to
adapt to a specific user.
Given the shortcomings described above, the subsequent work will include a
study of which gestures should be assigned in advance, which of them require
adjustment, with which gestures to perform functions attached to them at the
end of the gesture, with which – at the beginning, and which gestures should be
used for visual display of the function execution process.
The proposed adaptation model can be used for access control, where the
system will select the most appropriate visualization model for the situation,
allow columns and rows of matrices to be sorted by user-friendly gestures, and
scale trees. The model is also applicable for controlling self-organizing sensor
networks, where, in accordance with the situation, a decision will be made to
display the network using a graph, TreeMaps, Voronoi maps, or some combina-
tion of them. Also, gestures will be selected that are most appropriate for the
selected visualization models.
In general, the approach allows one to speed up decision-making processes
and improve their quality when setting up mobile device networks. For example,
when analyzing networks, the use of gestures allows one to quickly and more
intuitively switch between metrics, capture the interested vertices or groups of
vertices, and switch between graph representations. Gestures can also be used
to manage access control (for example, when managing permissions between
mobile devices) and to assess risks (for example, when assessing the risk and
cost of losing a device). Separately, it is worth noting the value of gestures when
used on tablets and mobile devices which is in demand in production – when

Adaptive Touch Interface: Application for Mobile Internet Security 71
a specialist needs to configure mobile device networks in the field. Thus, the
approach also expands the possibilities of using visual analytics for situations
when using a PC is difficult.
6 Conclusion
The paper proposes the approach to human-computer interaction with the inter-
faces of mobile internet security applications based on touch screens.
The paper proposes the models of user interaction and visualization, the
adaptation algorithm and the “best practices” gesture design algorithm. Exper-
iments on the perception of gestures by users on the examples of visual analytics
for the hierarchical centralized network of embedded devices and the decentral-
ized sensor network were carried out. The methodology proposed in this paper
can be used to create new models of interaction with the touch interface in the
risk assessment process.
Further research will be aimed at studying the naturalness of gestures on
touch screens in the perception of users, as well as studying the best fit of
gestures to the visual display of information security metrics.
References
1. Apple — use multi-touch gestures on your Mac - Apple support. https://support.
apple.com/en-us/HT204895. Accessed June 2019
2. Apple developer — gestures - user interaction - IOS - human inter-
face guidelines. https://developer.apple.com/design/human-interface-guidelines/
ios/user-interaction/gestures/. Accessed June 2019
3. Apple developer — UI design dos and donts. https://developer.apple.com/design/
tips/. Accessed June 2019
4. kolenda, N.: Psychology and business. https://www.nickkolenda.com/user-
experience/#. Accessed June 2019
5. World Usability Congress: Touch screen usability best practices when designing
automation user interfaces (UI). https://worldusabilitycongress.com/touch-screen-
usability-best-practices-when-designing-automation-user-interfaces-ui/. Accessed
June 2019
6. Ananto, B.S., Sari, R.F., Harwahyu, R.: Color transformation for color blind com-
pensation on augmented reality system. In: 2011 International Conference on User
Science and Engineering (i-USEr), pp. 129–134. IEEE (2011)
7. Best, D.M., Bohn, S., Love, D., Wynne, A., Pike, W.A.: Real-time visualization
of network behaviors for situational awareness. In: Proceedings of the 7th Interna-
tional Symposium on Visualization for Cyber Security, pp. 79–90. ACM (2010)
8. Bishop, M.: Conspiracy and information flow in the take-grant protection model.
J. Comput. Secur. 4(4), 331–359 (1996)
9. Choi, H., Lee, H., Kim, H.: Fast detection and visualization of network attacks on
parallel coordinates. Comput. Secur. 28(5), 276–288 (2009)
10. Desnitsky, V., Levshun, D., Chechulin, A., Kotenko, I.V.: Design technique for
secure embedded devices: application for creation of integrated cyber-physical secu-
rity system. JoWUA 7(2), 60–80 (2016)

72 K. Zhernova et al.
11. Heitzmann, A., Palazzi, B., Papamanthou, C., Tamassia, R.: Effective visualization
of file system access-control. In: Goodall, J.R., Conti, G., Ma, K.-L. (eds.) VizSec
2008. LNCS, vol. 5210, pp. 18–25. Springer, Heidelberg (2008). https://doi.org/10.
1007/978-3-540-85933- 8 2
12. Hutchins, E.L., Hollan, J.D., Norman, D.A.: Direct manipulation interfaces. Hum.
Comput. Interact. 1(4), 311–338 (1985)
13. Ingols, K., Lippmann, R., Piwowarski, K.: Practical attack graph generation for
network defense. In: 2006 22nd Annual Computer Security Applications Confer-
ence, ACSAC 2006, pp. 121–130. IEEE (2006)
14. Kim, D.-K., Ray, I., France, R., Li, N.: Modeling role-based access control
using parameterized UML models. In: Wermelinger, M., Margaria-Steffen, T. (eds.)
FASE 2004. LNCS, vol. 2984, pp. 180–193. Springer, Heidelberg (2004). https://
doi.org/10.1007/978-3-540-24721- 0 13
15. Kolomeec, M., Chechulin, A., Kotenko, I.V.: Methodological primitives for phased
construction of data visualization models. J. Internet Serv. Inf. Secur. 5(4), 60–84
(2015)
16. Kolomeec, M., Chechulin, A., Pronoza, A., Kotenko, I.V.: Technique of data visu-
alization: example of network topology display for security monitoring. JoWUA
7(1), 58–78 (2016)
17. Kolomeec, M., et al.: Choosing models for security metrics visualization. In: Rak,
J., Bay, J., Kotenko, I., Popyack, L., Skormin, V., Szczypiorski, K. (eds.) MMM-
ACNS 2017. LNCS, vol. 10446, pp. 75–87. Springer, Cham (2017). https://doi.org/
10.1007/978-3-319-65127- 9 7
18. Kolomeets, M., Chechulin, A., Kotenko, I., Saenko, I.: Access control visualization
using triangular matrices. In: 2019 27th Euromicro International Conference on
Parallel, Distributed and Network-Based Processing (PDP), pp. 348–355. IEEE
(2019)
19. Kolomeets, M., Chechulin, A., Kotenko, I., Strecker, M.: Voronoi maps for pla-
nar sensor networks visualization. In: You, I., Chen, H.-C., Sharma, V., Kotenko,
I. (eds.) MobiSec 2017. CCIS, vol. 971, pp. 96–109. Springer, Singapore (2019).
https://doi.org/10.1007/978-981-13- 3732-1 7
20. Kotenko, I., Doynikova, E.: Security assessment of computer networks based on
attack graphs and security events. In: Linawati, M.M.S., Neuhold, E.J., Tjoa, A.M.,
You, I. (eds.) ICT-EurAsia 2014. LNCS, vol. 8407. Springer, Heidelberg (2014).
https://doi.org/10.1007/978-3-642- 55032-4 47
21. Kotenko, I., Stepashkin, M., Doynikova, E.: Security analysis of information sys-
tems taking into account social engineering attacks. In: 2011 19th International
Euromicro Conference on Parallel, Distributed and Network-Based Processing, pp.
611–618. IEEE (2011)
22. Kotenko, I., Ulanov, A.: Simulation of internet DDoS attacks and defense. In:
Katsikas, S.K., L´opez, J., Backes, M., Gritzalis, S., Preneel, B. (eds.) ISC 2006.
LNCS, vol. 4176, pp. 327–342. Springer, Heidelberg (2006). https://doi.org/10.
1007/11836810 24
23. Roberts, J.C.: Guest editor’s introduction: special issue on coordinated and mul-
tiple views in exploratory visualization. Inf. Vis. 2(4), 199–200 (2003)
24. Sarkar, M., Brown, M.H.: Graphical fisheye views. Commun. ACM 37(12), 73–83
(1994)