202
Цифровые технологии в системе уголовно-правовых отношений
А. Е. Середа,
ведущий специалист,
учебная лаборатория криминалистической техники и судебных экспертиз,
кафедра криминалистики юридического факультета,
Белорусский государственный университет
ИСКУССТВЕННЫЙ ИНТЕЛЛЕКТ В КРИМИНАЛИСТИЧЕСКОМ
ИСПОЛЬЗОВАНИИ ТЕХНОЛОГИЙ АВТОМАТИЧЕСКОГО
РАСПОЗНАНИЯ ЛИЦ
Аннотация.
В статье рассматриваются аспекты применения искусственного
интеллекта в автоматических системах распознания лиц в рамках поддержания
национальной безопасности. Также рассмотрению подвергается такая особенность
искусственного интеллекта как «глубокое обучение» и его роль в криминалистиче-
ском обеспечении расследования преступлений.
Ключевые слова
: криминалистика, искусственный интеллект, глубокое об-
учение, алгоритмы распознания, распознание лиц, технологии, преступность,
национальная безопасность, криминалистическая идентификация, верификация
ARTIFICIAL INTELLIGENCE IN THE FORENSIC USAGE OF AUTOMATIC
FACIAL RECOGNITION TECHNOLOGIES
Abstract.
The article deals with the issues of ensuring national security through
deployment of facial recognition technologies in forensics. It also takes a close look at
the emerging trend of utilizing artificial intelligence in this field and aims to unravel the
mystery of the so-called Deep learning by focusing on the phenomena being incorporated
into facial recognition technologies.
Keywords
: Forensics, Artificial intelligence, Deep learning, Algorithms, Facial
recognition, Technology, Crime, National security, Identification, Verification
Introduction
. The issues of ensuring national security are now becoming increasingly
relevant affecting the interests of a wide range of subjects. The volume of information as
well as information technologies and the processes of globalization form the conditions
under which the world information arena becomes a means of achieving various goals,
namely a unique phenomenon and a new sphere of society’s day-to-day life. Taking into
account the increasing volume and availability of information and information resources,
the changing technological landscape of modern society, it seems necessary to take into
consideration the challenges and problems unique to our time requiring close attention
from the legislative and executive branches of the government.
On the international level, the problem of combating crime is quickly gaining
worldwide relevancy and significance. The world economy is being actively transformed and
is characterized by increased instability, caused in part by transition to a new technological
order. On a global scale, markets are being redistributed, financial flows and productive
forces are also being redistributed, and competition is becoming tougher through escalation
of interests. Technological evolution is rapidly becoming a source of fundamentally new
203
Цифровые технологии в системе уголовно-правовых отношений
threats and opportunities, providing previously inaccessible options for both negative
and positive influence on the individual, state, and society in general. Many indicators
of crime (dynamics, structure, latency, detection) largely depend on the effectiveness of
the fight against professional crime and its prevention, which cannot be fully conducted
without the use of modern technological advances.
Successful crime prevention in 2022 is characterized in large part by new strategies
and approaches to its implementation. It should be noted that one of the most promising
tools for the investigation, solving and prevention of crimes can be the technology capable
to strip a perpetrator of his actual or perceived (in case of crime prevention) anonymity.
In modern days, the main driving force of crime, in addition to the motive for committing
a crime (usually it’s greed, jealousy or hatred), is the expectation of the criminal that his
identity will remain unknown. Thus, a potential criminal, hiding his identity, eventually
commits a felony or misdemeanor (depending on the illegal activity)
–
as examples of
this could be seen the riots and property destruction by mobs in Minsk on August 20,
2020; uprising on the Capitol Hill in the USA on January 6, 2020; riots and damage to
property by the Black Lives Matter organization on May 25 in the USA; aggressive
protests by truckers on February 12, 2022 in Canada, etc. For the same reason, we can
observe an exponential increase in cybercrime around the world, namely perceived or
actual anonymity of a perpetrator.
The most effective way to deprive a criminal of his anonymity appears to be the
automatic face recognition technology (AFRT). However, in the Russian-language
literature, there is almost no analysis of the concepts of face recognition proposed by
various researchers
–
a clear classification of face recognition in the investigation of crimes
is not proposed; prospects for further implementation in the investigation of crimes in
the future are not offered. All these aspects are extremely important for the current stage
of the development of criminalistics.
Main div
. Throughout its development, face recognition in criminalistics has gone
through several stages of its development: 1) verbal portrait; 2) bertillonage-anthropometry;
3) photo-video technology; 4) automatic recognition by facial characteristics.
In turn, the process of development of automatic recognition by facial characteristics
and other signs can also be divided into the following stages of the formation of this
technology: 1) quasi-automatic computerized face recognition systems; 2) fully automatic
recognition and the emergence of biometric databases of persons for identification; 3) the
use of linear algebra to work with images; 4) the implementation of this technology to
ensure the safety of public order, disclosure and investigation of crimes; 5) the development
and use of numerous algorithms in the development of automatic face recognition systems;
6) 3D face recognition and artificial intelligence [1. P. 16–17].
Over the past 10 years, the use of automatic face recognition technology in video
surveillance systems has become increasingly common worldwide. In total, there are
currently 109 countries that either use or have approved the use of facial recognition
technology for surveillance purposes. Since the development of automatic face recognition
systems, much attention has been paid to its further progress and implementation into
the technological and scientific landscape of society by such foreign authors as Mark
Kempster, Alberink, Ruyfork, Desmoz, Champode, Messer, Harley, Chunk, etc.
204
Цифровые технологии в системе уголовно-правовых отношений
Kempster considers face recognition in the context of law enforcement and gives the
following definition: «Face recognition is an activity aimed at assimilating information
that is hidden in the general set of facial features of an individual. This information can be
read and recorded by giving mathematical and graphic design of the presented features and
characteristics of the person, which, in their separate state, do not represent information
significant for identification of the individual, but in their totality, are the business card of
the individual». According to this point of view, face recognition is a unifying process,
which consists in giving a system to separately presented features, which together represent
identifying information. This approach is extremely common and it was this point of view
that formed the basis of the first semi-automatic face recognition systems [1. P. 31–32].
Alberic and Ruyfork see face recognition as a process of presenting the received
information in a graphical form and further identifying a person, drawing an already thin
line between the two concepts. Face recognition, in this case, is the process of identifying
the correspondence of the characteristic features and features inherent in the face to the
already available and graphically presented characteristics of the face for comparison.
Simply put, this point of view reduces two different processes: recognition and identification
of a person to a single process [5, P. 42–43].
Dezmoz positions face recognition as a process of fixing and mathematically
displaying static and dynamic characteristics of a face. This approach is due to the fact
that the researcher sees identifying signs in such dynamic characteristics as blinking,
lip movement, nervous twitching or rolling of the eyes, narrowing and widening of the
nostrils, etc. «Face recognition», stipulates the researcher, «is not possible without fixing
and displaying all signs without exception: including dynamic ones, which, for the most
part, are unique to humans» [5. P. 46–49].
In the position of the researcher, it is clearly expressed that he operates from the point
of view of the latest technologies
–
automatic face recognition systems. We believe that the
recognition of dynamic facial features seems to be an appropriate aspect of face recognition
and, of course, will be of great interest for the detection and investigation of crimes.
Champod, developing his approach in collaboration with Dezmoz, proposed his own
classification of signs for recognition: 1) signs of belonging
–
belonging to race, nation, gender,
etc. 2) identifying (or unique) signs
–
forehead height, eye color, mouth width, scars, etc.
Further, the researcher recommends dividing the signs into 1) general
–
the largest elements,
such as: figure, clothing size, hair color; 2) private
–
these are components of the general signs
that their detail; 3) permanent signs (define individual traits); 4) temporary (tan, tattoos, hair,
teeth, etc.); 5) accidental (traces of diseases, skin pigmentation, acne) [6. P. 87–88].
Messer and Harley argue that face recognition is always a semi-automatic process,
carried out not without the participation of a specialist in the respective field and computer
technology (in the case of face recognition, biometric databases and the algorithm used in
the recognition process). In recognition using modern technologies, researchers distinguish
3 main stages: 1) preliminary
–
reading and fixing of information; 2) intermediate
–
loading
of the received information into the database for identity search; 3) final
–
control verification
of the result by an expert. Obviously, Messer and Harley took as a basis the position of
Alberink and Ruyfork that face recognition includes the final identification of a face
–
the
difference is that they focused on systematization of the process, highlighting the need for
the participation of a specialist at the final stage of face recognition [8. P. 277–280].
205
Цифровые технологии в системе уголовно-правовых отношений
Chunk pointed out the need to include in the concept of «Face recognition» such
an element as a dynamic change in the features of the external appearance of the face,
its characteristic properties and features in the aging process. At its core, the researcher,
taking into account the technological development and the increase in the capabilities
of modern algorithms for automatic face recognition, expands the concept itself. Face
recognition, according to him, not only consists of fixing and processing the data available
at the time of reading the information, but also the use of artificial intelligence, whose task
is to predict how much the appearance of this person will change in the future, or how
his face looked in the past. Thus, according to Chunk, it will be possible to uncover and
investigate crimes of considerable prescription, as well as to predict in advance what a
potential criminal will look like in the future, which will greatly simplify the achievement
of the ultimate goal of face recognition –identification of the criminal [8. P. 211–212].
According to the American criminology researcher Richard Seiferstein, facial
recognition technology is a form of biometric artificial intelligence (AI) that performs
identity verification by comparing video frames or digital images and matching them with
images of faces stored in a database based on facial features and skin texture. It provides
automatic, fast and unhindered verification, since no physical contact, such as fingerprints
or other security measures, is required. In addition, it is not related to any keys or identity
cards that may be stolen or lost.
At the current stage of development, researchers agree that face recognition in the
form in which it exists as of 2022 is a revolutionary technology, one of the important
components of which is Deep learning.
Deep learning is a type of Machine learning and artificial intelligence (AI) that
simulates how people acquire certain types of knowledge. Deep learning is an important
element of data science, which includes statistics and predictive modeling. In its simplest
form, Deep learning can be considered as a way to automate predictive analytics. While
traditional Machine learning algorithms are linear, Deep learning algorithms are built
into a hierarchy of increasing complexity and abstraction.
In traditional Machine learning, the learning process is controlled, and the programmer
must be extremely precise, telling the computer what types of objects he should look for
in order to decide whether an image contains a particular object. This is a time-consuming
process called «Function extraction», and the success of the computer depends entirely on
the programmer’s ability to accurately determine the set of functions. The advantage of Deep
learning is that the program itself creates a set of functions without human supervision [2].
Investigating the issues related to the use of technical and forensic tools in the
context of general aspects of fixing evidence in criminal proceedings, scientists note
that the tools used are not just auxiliary in working with evidence, contributing to their
capture and preservation. For example, in accordance with the opinion of A. A. Levy, video
recording with face recognition «performs an essential cognitive function, to a certain
extent replaces the court’s direct perception of recorded information»
However, face recognition may have a different definition depending on the method,
object and purpose of recognition. Van der Lugt, a researcher of automatic face recognition
systems, pointed out that face recognition by a person, technology with human participation, or
206
Цифровые технологии в системе уголовно-правовых отношений
fully automatic face recognition does not fall under the same category of face recognition and,
therefore, should be considered different in its essence and nature processes [10. P. 115–118].
In the same vein, David Kuhn emphasizes, we can talk about the nature of the activity
in which facial features and characteristics are recognized. The researcher adheres to a certain
dichotomy in this matter: the recognition of a face, its dynamic and static features, can be
carried out within the framework of law enforcement activities (investigation of crimes,
as well as their prevention), as well as within the framework of maintaining security (face
recognition in electronic devices, for admission to a limited territory, etc.) [11. P. 65–68].
In addition, it also seems to us extremely important to highlight a distinction between
the concepts of «Identification» and «Verification»
–
both concepts are important to the
study, application and legislative consolidation of automatic face recognition technology,
since, as a result of the study of automatic face recognition technology, at the present
stage, the result of face recognition is always one of the following: face verification,
identification, or recognition of emotions.
Verification of a person, in this regard, is the result of the congruence of the biometric
characteristics of the person captured in the photo with the characteristics of the person stored in
the database. In other words, the comparison and search for matches of biometric characteristics
of the same person in the photo and in the available database is carried out. In case of positive
verification, access to a place is provided, a service is provided, etc. [7. P. 96–97].
Identification of a person is fundamentally different from verification. Identification
of a person by biometric characteristics of a person consists in finding matches of the
facial portrait compiled by the algorithm of the face recognition program with numerous
others presented in the database. We are talking about establishing the identity of a suspect,
victim, accused or other participant in the criminal process, as well as missing persons
or victims of an accident [7. P. 91–92].
As of 2022, the main array of automatic face recognition systems consists of 2D (two-
dimensional) recognition systems. The 2D facial recognition technology is based on flat
two-dimensional images. Face recognition algorithms use: anthropometric parameters of
the face, graph models of faces or elastic 2D models of faces, as well as images with faces
represented by a certain set of physical or mathematical features. 2D image recognition is
one of the most popular technologies at the moment. Since the main databases of identified
persons accumulated in the world are precisely two-dimensional, the main equipment
already installed around the world is predominantly 2D.
However, the development and implementation of 3D scanning and facial recognition
systems is already underway. 3D recognition is usually performed in a reconstructed
three-dimensional way. This type of facial recognition technology has higher quality
characteristics. There are several different 3D scanning technologies: these can be
laser scanners with an estimate of the range from the scanner to the elements of the
object’s surface, special scanners with structured illumination of the object’s surface
and mathematical processing, or they can be scanners that process photogrammetric
synchronous stereo pairs of images of faces.
Automatic face recognition includes several stages [4]:
1. Face detection. The camera will detect a person’s face, whether he is alone or
in a crowd. The face is more effectively detected at the moment when a person looks
207
Цифровые технологии в системе уголовно-правовых отношений
directly into the camera; however, modern technological advances make it possible to
detect a face in situations when a person is not looking directly (within certain limits).
2. Face analysis. When a photo of the face has been taken, its analysis begins. Most
face recognition solutions use 2D images instead of 3D volumetric images, because they
can more easily match 2D photos with publicly available photos or photos available in a
database. Each face is made up of distinguishable landmarks or nodal points (such as the
distance between the eyes or the shape of the cheekbones). Face recognition programs
analyze these nodal points in depth.
3. Converting images to data. After that, the result of analysis of the face turns into
a mathematical formula. Facial features become a numeric code. Such a numeric code
is called a «Faceprint». Like the unique structure of a thumbprint, each person has their
own «Faceprint».
4. Search for matches. Then the received code is compared with photos with IDs
in the database.
Modern face recognition technologies, namely their effectiveness is measured by
two indicators [5. P. 114–115]:
1) The level of erroneous confirmations (hereinafter referred to as FAR) is the
probability that the facial recognition system mistakenly identifies an unregistered user
or confirms his authenticity
2) The level of erroneous failures (hereinafter referred to as FRR) is the probability
that the system does not identify the registered user or does not confirm his authenticity.
The FRR calculation formula looks like this:
Where is the number of image standards in the database. FR is the number of false
non-recognitions.
FAR is calculated similarly
Where is the number of image standards in the database. FA is the number of false
recognitions.
Table 1 shows these indicators for automatic face recognition systems in 2D and
3D format. Based on these indicators, it is preferable to choose a system for certain
purposes. Of course, it seems rational to opt for the system with the best indicators
(table 1) [5. P. 120–122].
Table 1. Comparison of FAR and FRR indicators for automatic face recognition
systems in 2D and 3D format
Approach
False pass ratio
(FAR)
False failure rate
(
FRR
)
3D face recognition
0.0005 %
0.1 %
2D face recognition
0.1 %
2.5 %
208
Цифровые технологии в системе уголовно-правовых отношений
The first and most important thing to note is that the indicators given in the table
are not absolute, but relative, i. e. they may vary depending on the settings of the facial
recognition algorithm. The second is that these indicators are interrelated
–
the smaller the
FAR, the greater the FRR. The approximate values of FRR and FAR for facial recognition
systems and their relationship are presented below (table 2) [5. P. 125].
Table 2. The relationship of the FAR and FRR coefficients for
automatic face recognition systems
FAR
FRR
0.1 %
2.5 %
0.01 %
7 %
0.001 %
10 %
The main disadvantages of 3D facial recognition technology include the following
aspects [5. P. 171]:
1) 3D recognition requires special cameras for scanning, which are several times
more expensive than conventional CCTV cameras that are used in 2D recognition;
2) lack of ready-made databases of identified persons, compared to 2D recognition;
3) recognition of twins remains a difficult task for facial recognition algorithms.
On average, 13.1 twins per 1000 newborns are born in the world, and this figure varies
greatly depending on the geographical region.
As of 2022, forensic facial recognition systems are used to track criminals and
identify wanted or missing persons. To demonstrate the full breadth of the scale, several
examples should be given [2, 3, 4]:
1) By the end of 2018, there were about 4,000 surveillance cameras installed in
Minsk as part of the Republican Public Security Monitoring System. In order to maintain
law and order, a Synesis product called «Kipod» is used, which allows setting up a
biometric system for identification and recognition of persons, as well as identification
and recognition of vehicle numbers;
2) In Moscow, as part of the Safe City program, one of the world’s largest face
recognition networks operates
–
more than 200 thousand video surveillance cameras.
This contributed to the disclosure of more than 5 thousand crimes in 2020. In addition,
it was found that the effectiveness of video surveillance systems in solving crimes has
an annual increase of 15–16 %;
3) In the People’s Republic of China, as of 2018, 170 million surveillance cameras were
installed and put into operation, and in the period from 2018 to 2021, another 400 million
units were installed. The «Dragonfly Eye» facial recognition system is used to maintain
public order. In the first three months of using this technology in Shanghai, law enforcement
officers detained 567 criminals, and the level of pickpocketing in the cities where it was
implemented fell by almost a third. The Zhengzhou City police, for example, use glasses
with a face recognition system that give out a person’s name and address in 2–3 minutes.
4) In the United States of America, the facial recognition system «FACES» (a system
for comparing and examining faces) is used, which is based on algorithms that scan
more than 30 million images from driver’s licenses and photos. As of 2022, out of 24 US
agencies, 18 already use facial recognition technologies, some use more than one system.
209
Цифровые технологии в системе уголовно-правовых отношений
In 2014, Facebook (The Facebook in Russia is recognized as extremist and banned.)
launched the «DeepFace» service, which determines whether two photographed faces
belong to the same person with an accuracy of 97.25 %. In 2015, Google introduced its
development
–
«FaceNet», which achieved a record accuracy of 99.63 % due to the huge
array of data collected by Google services. The technology, in particular, is used in «Google
Photos» to sort images and automatically mark people on them [3].
On September 12, 2017, Apple introduced «Face ID» technology, replacing the
fingerprint sensor «Touch ID». The technology developed by Apple is unique in the sense
that it contains the following elements: 1) a dot projector
–
it projects more than 30,000
invisible infrared dots onto the user’s face, using which his mathematical model is then
created; 2) an infrared camera
–
reads the point structure of the face, creates an image in
the infrared spectrum and places this data into a special processor module; 3) infrared
emitter
–
emits an invisible beam of infrared light on the face, which allows one to perform
an accurate scan of it even in complete darkness [3].
In addition, companies such as Clearview AI, Vigilant Solutions and Acuant FaceID
are also working on their face recognition systems. The enormous amount of information
collected by the private sector can be useful for law enforcement agencies, since government
agencies have many channels of access to corporate data. In 2020, from January to June
alone, federal, state and local law enforcement agencies in the United States sent more than
112,000 legal requests for data to Apple, Google, Facebook and Microsoft
–
three times more
requests than in 2015 (of which approximately 85 % were accepted and responded to) [9].
According to the study «Facial Recognition Market», there are the following
algorithms for automatic face recognition (table 3) [3]:
Table 3. The most common, as of 2022, algorithms of automatic face recognition
systems in the world
Algorithm
Manufacturer
Country of origin
megvii-000
Megvii
China
visionlabs-003
VisionLabs
Russia
visionlabs-002
VisionLabs
Russia
morpho-002
Morpho
France
morpho-000
Morpho
France
ntechlab-003
NtechLab
Russia
ntechlab-002
NtechLab
Russia
cogent-000
Gemalto Cogent
USA
vocord-002
Vocord
Russia
fdu-000
Fudan University
China
fdu-001
Fudan University
China
neurotechnology-003
Neurotechnology
Lithuania
itmo-003
ИТМО University
Russia
3divi-001
3DiVi Inc.
Russia
yitu-000
Yitu Technologies
China
gorilla-000
Gorilla Technology
Taiwan
cyberextruder-002
CyberExtruder
USA
210
Цифровые технологии в системе уголовно-правовых отношений
As of 2022, the regional application of automatic face recognition technology is as
follows [3]:
1) Half of North American countries currently use automatic facial recognition
technology (50 %);
2) In South America, the vast majority of countries use automatic face recognition
technology (92 %);
3) The countries of the Middle East and Central Asia largely use the technology
of automatic face recognition (76 %);
4) More than half of European countries currently use automatic face recognition
technology (69 %);
5) The smallest number of countries use facial recognition technology in the
investigation of crimes (20 %).
Conclusion
. The concept of face recognition is quite deep for the reason that
it implies a multi-level structure, each element that corresponds or corresponded to
the truth at a particular stage of technology development. Currently, based on modern
research and the works of authors developing this issue, the following seems to us to be
the correct definition of this concept: «Face recognition in criminalistics is an automatic,
semi-automatic or manual process aimed at assimilation of information (static or dynamic
features) that is hidden in the total set of facial features of an individual. This information
can be read and recorded by giving mathematical and graphic design of the presented
features and characteristics of the person, which, in their separate state, do not represent
information significant for identification of the individual, but in their totality, are a reliable
identifier of the individual».
As of 2022, face recognition systems begin to deploy AI algorithms and Machine
learning (also known as Deep learning) to detect human faces. The algorithm typically
starts by searching for human eyes, followed by eyebrows, nose, mouth, nostrils, and
iris. Once all the facial features are captured, additional validations using large datasets
containing both positive and negative images confirm that it is a human face. Some of
the common techniques used for facial recognition are feature-based, appearance-based,
knowledge-based, and template matching. Each of these methods has its advantages and
disadvantages.
Feature-based methods rely on features such as eyes or nose to detect a face. The
outcomes of this method could vary based on light. Further, appearance-based methods
use statistical analysis and Machine learning to match the characteristics of face images.
In a knowledge-based approach, a face is recognized based on predefined rules. Template-
matching methods compare images with previously stored face patterns or features and
correlate the results to detect a face. However, this method fails to address variations in
scale, pose, and shape.
Machine learning/Deep learning is a subset of AI that mainly focuses on using data
and algorithms to mimic human natural learning processes. It uses statistical methods to
train algorithms to classify or predict and even provide insights into data mining projects.
Terms like Deep learning and Machine learning and sometimes neural networks are used
in the industry interchangeably. However, there are subtle differences between these
technologies. A neural network is a subset of Deep learning while Deep learning is one
211
Цифровые технологии в системе уголовно-правовых отношений
of the arms of Machine learning. Simply put, Deep learning involves training algorithms
with minimal human intervention. It converts unstructured data to manageable groups
for processing through a process known as dimensionality reduction.
On the other hand, neural networks also known as artificial neural networks comprise
node layers
–
an input layer, multiple hidden layers, and an output layer. Each of the nodes
has an associated weight and threshold and is connected to the other nodes. Basically,
if the value of any output layer exceeds its threshold, data is sent to the next layer of
the network. Neural networks are of two types: basic neural networks and Deep neural
networks. In the basic neural network, two or three layers are present whereas a deep
neural network consists of more than three layers.
Artificial Intelligence and Machine learning offer a multitude of opportunities and
endless possibilities to work for the betterment of the world. However, it is essential to
pay attention to the ethics and privacy of people while dealing with data. Data storage,
management, and security are the other aspects that will play an important role in making
these technologies invasive. In order to overcome the problematic aspects of the use of
facial recognition technology in the detection and investigation of crimes that are relevant
as of 2022, it is necessary:
1. Improving the accuracy and reliability of facial recognition systems in the detection
and investigation of crimes. To achieve this goal, it is advisable to implement:
– the installation of a larger number of video surveillance cameras, with a high
resolution of video recording, a large amount of memory and capable of recording in low
light conditions or bad weather. In addition, a single format for recording information
should be established;
– development of algorithms for recognizing persons who work equally effectively
with representatives of different races and nationalities.
2. Modernization of the privacy and data security infrastructure, which will require:
– installation or development of software that will reliably protect the received data;
– the development of a regulatory legal act of the concept of protection and processing
of biometric data providing for responsibility for failure to ensure security in this matter.
3. Development of relevant regulatory legal acts regulating the use of facial
recognition technology in integration with existing databases. It is necessary to develop
a legal framework that will regulate in detail the use of the interaction of these systems
in the investigation and prevention of crimes.
4. Development of training programs for specialists in the field of facial recognition.
It is of paramount importance to develop and ultimately deploy training programs for
specialists in the field of working with up-to-date face recognition technologies and devising
new, more advanced algorithms. As well as there is forensics and counter-forensics, face
recognition systems meet their opponents in the form of various kinds of tricks and fakes
(for example, the notorious «Deepfake»), as well as means and methods of hiding the
characteristic properties of appearance. Up-to-date knowledge of specialists working
with facial recognition systems and developers of new algorithms for these systems will
allow investigating and solving crimes most accurately and with minimal time costs.
5. Adaptation of algorithms to changes in the appearance of the aging process, as
well as surgical interventions and other deformities of the face. The variability of the
signs of the suspects, for the most part, implies age-related and surgical changes.
212
Цифровые технологии в системе уголовно-правовых отношений
Taking into account the fact that some changes can radically transform a person’s
appearance, the following is advisable:
1) creation of data banks related to gender, age and ethnicity which provide
background information for a variety of diagnostic, clinical and judicial procedures;
2) development and use of an algorithm that is able to take into account the
approximate age change of a person after a certain time.
Elements with a bone base, such as the skull, forehead, do not undergo serious changes
throughout a person’s life. However, the nature of their transformation can mainly be
assessed by experts with significant experience or specially developed software algorithms.
References
1. Colmenarez, A. Facial Analysis from Continuous Video with Applications to
Human-Computer Interface / A. J. Colmenarez, Z. Xiong, T. S. Huang.
–
Kluwer Academic
Publishers, New York, Boston, 2004. – 159 p.
2. Face off. Law Enforcement Use of Face Recognition Technology. URL:
https://www.eff.org/files/2019/05/28/face-off-report.pdf (дата обращения: 11.07.2022).
3. Facial Recognition Market Size, Share & Trends Analysis Report by Technology
(2D, 3D, Facial Analytics), by Application (Access Control, Security & Surveilance),
by End-use, by Region, and Segment Forecasts, 2021–2028. URL: https://www.
grandviewresearch.com/industry-analysis/facial-recognition-market (дата обращения:
14.07.2022).
4. Guidelines 05/2022 on the use of facial recognition technology in the area of law
enforcement. URL: https://edpb.europa.eu/system/files/2022–05/edpb guidelines_202205_
frtlawenforcement_en_1.pdf.
(дата обращения: 12.07.2022).
5. Lambert, W. Issues with Facial Recognition Technology / W. Lambert.
–
Nova
Science Publishers inc., 2022. – 232 p.
6. Lee-Morrison, L. Portraits of Automated Facial Recognition. On Mechanic Ways
of Seeing the Face / L. Lee-Morrison.
–
Majuskel Medienproduktion GmbH, Werzlar,
2019. – 199 p.
7. Lyle, D. Forensics / D. Lyle.
–
Fraser Direct, 100 Armstrong Avenue, Georgetown,
Canada, 2008. – 494 p.
8. Rattani, A. Selfie Biometrics. Advances and Challenges / A. Rattani, R. Derakhshani,
A. Ross.
–
Zillow inc., Seattle, 2020. – 377 p.
9. Racial Discrimination in Face Recognition Technology. URL: https://sitn.hms.
harvard.edu/flash/2020/racial-discrimination-in-face-recognition-technology/ (дата об-
ращения: 12.07.2022).
10. Saferstein, R. Criminalistics. An introduction to Forensic Science / R. Saferstein,
R. Tiffany.
–
Mt. Laurel, New Jersey, 2020. – 577 p.
11. Vatsa, M. Deep Learning in Biometrics / M. Vatsa, R. Singh, A. Majumdar.
–
Taylor&Francis Group, LLC, 2018. – 329 p.
