See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/366394644

MAIN AREAS OF APPLICATION OF ARTIFICIAL INTELLIGENCE BY CRIMINAL

GROUPS

Chapter · December 2022

CITATIONS READS

0 69

1 author:

Paul Neumann
University Metropolitan Tirana
31 PUBLICATIONS   2 CITATIONS   

SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Digital Criminology View project

Omerta.com View project

All content following this page was uploaded by Paul Neumann on 18 December 2022.

The user has requested enhancement of the downloaded file.

 MAIN AREAS OF APPLICATION
OF ARTIFICIAL INTELLIGENCE
BY CRIMINAL GROUPS
Paul Neumann
Digital Criminology

n 2018, the FBI hosted an international conference on cyber security, which ad-

I dressed the issues of the use of artificial intelligence by law enforcement authori-
ties and criminals. The conference noted: data from Interpol, Europol, FBI, and law
enforcement agencies of other countries, as well as research results from leading uni-
versities, suggest that there are currently no signs of purposeful efforts of organized
crime to pursue their own developments in the field of artificial intelligence.
Many cyber security experts interpret it in the way that in the short term, the FBI
and the police, using artificial intelligence, will gain a decisive superiority over cyber
crime and other types of organized crime. Successful criminals working at large in
such areas as finance, trafficking, intellectual property, etc., are extremely rational peo-
ple. At this level of development in the field of artificial intelligence, they do not need
to attract attention by recruiting teams of the most advanced start-ups that are always
in sight of the military and intelligence communities, as well as large corporations.
Currently, is not needed, and not without a reason.
First of all, in an effort to minimize costs and attract the maximum number of ex-
ternal, largely unemployed, developers to the development of their own product, most
of the leading manufacturers of artificial intelligence platforms have already released
open-source platforms. Cyber criminals have plenty to choose from to build their own
powerful AI platforms. Almost all open-source artificial intelligence developments are
containers. A container is a platform on which, using the API, any third-party pro-
grams, services, databases, etc. can be mounted. If in the past every developer had
to start creating his platform or service from building an algorithm, and translating it
into a code using a programming language, then now it is possible to create products
and services the same way as builders build a house – from standard components
delivered to the construction site.
Since 2016, the artificial intelligence as a service (AIAS) concept has been growing
rapidly. Companies that develop individual elements of artificial intelligence, first of
all, data warehouses, deep learning algorithms, neural network algorithms, including
deep ones, as well as natural language processing and multi-dimensional computing
programs, in more than two-thirds of cases provide for the possibility of using their
developments through the API. Moreover, some companies nowadays allow for a rel-
atively small fee to rent their AI software. If in the case of API, if necessary, law
enforcement experts can establish the nature of the use of the program, then renting
makes this impossible.

1
Digital Criminology

Finally, just a few years ago, there were 17 universities in the world, where students
received first-class training in research and practical development related to artificial
intelligence. In those days, law enforcement could easily keep an eye on every person,
who specialized in the field of an increased level of threat to the society, and could
track his career throughout his life. Now, this opportunity no longer exists. In the
United States alone, the number of universities teaching computer science at the top
level has increased to 40, and similar educational institutions have sprung up around
the world. An entire on-line learning industry has emerged. Today, for people with
the necessary initial training, the best universities have opened free on-line courses on
all components of artificial intelligence.
It is being estimated that more than 70% of AI development is made in the United
States, and about 15% in China. Accordingly, the main examples of the use of artificial
intelligence by criminals can be gleaned from reports and other sources published by
US law enforcement agencies (Chinese sources are strictly classified). The foregoing
facts indirectly indicate the active preparation of criminals to master artificial intelli-
gence. At the same time, the criminals are not going to reinvent the wheel. They
are more concerned with learning how to ride it, and work out the most efficient
routes. According to Interpol and the FBI, the use of artificial intelligence by crimi-
nals in America and other developed countries over the next five years will take place
in several priority areas. They are united by the ratio of three variables that is most
favourable for the crime: the criminal income received, the reduction of combined
costs of preparation, committing and concealing the crime, and the level of risk.
The use of artificial intelligence by criminal organizations boils down to four main
areas of application:

1. The use of artificial intelligence to compromise and implant malicious software

into payment systems that mainly use the blockchain protocol, and have a P2P
architecture.
Peer-to-peer payment systems are replacing processing companies, primarily
due to cost savings for customers. At the same time, as of 2019, out of nearly
30 payment services built on the blockchain operating in the United States, only
seven met the requirements of computer security. Accordingly, connecting to
payment services and adding about 0.1-0.3% to each transaction will bring bil-
lions of dollars in income to criminals without any risk. Artificial intelligence
programs are extremely important in this case. They allow the use of deep learn-
ing techniques for neural networks to hack and reprogram payment protocols
built on the blockchain. Research experiments have shown that artificial intelli-
gence programs are more effective at this task than human programmers. The
vulnerability lies in the blockchain. Like any code, it is based on rules and algo-
rithms. It is on them that games are built – from chess to poker, where artificial
intelligence rivals humans.

2. High-tech cyber criminals that profit from the trading operations of the largest
financial institutions account for 40-50 billion US dollars annually. It is the most
lucrative, albeit risky, area of organized cyber crime.
Since a regular financial arms race has unfolded in the last few years, manifested

2
 Main areas of application of artificial intelligence by criminal groups

through the improvement of IT platforms based on artificial intelligence at all

the largest financial institutions, criminals, in order to at least keep a share of
the income, need to participate in this race. In this regard, organized crime’s
use of artificial intelligence in operations in financial markets for penetrating
and compromising trading platforms leaves no other opportunity for criminals
but to use the best open source solutions combined with AIAS. In contrast to
the situation in the payment business, where a sharp increase in the size and
proportion of the criminal share in the turnover of payment systems was noted
in 2017-2020, in the algorithmic trading, in the short term, the proportion of
criminal gains will decrease.

3. There is reason to believe that as the technology race unfolds, the interest of
cyber criminals both inside and outside the United States in intellectual property
will only grow. It is also known that multi-functional programs based on self-
improving algorithmic modules are increasingly used to open today’s powerful
corporate information security systems. Such modules are a key element of
artificial intelligence. The FBI officials themselves believe that America today is
not ready to fight back hacker groups targeting intellectual property belonging
to the federal government, corporations, and universities.
In the technological race, the introduction of new products, services, and pro-
grams is an imperative to survival. It is clear that it is much more profitable to
buy stolen documentation, blueprints, or programs than to spend huge amounts
of money on research and development. A joint study in India by the FBI
Academy and pharmaceutical giant Sanofi S.A. showed that each $1 spent on
stolen intellectual property in pharmaceuticals saves generic manufacturers $17-
20 in research and development costs.
Obviously, such a lucrative industry is one of the key candidates for the use
of artificial intelligence. There is evidence that at present, within the corporate
networks of the leaders of the American high-tech and biotechnology, there are
multi-purpose and multi-functional hacker software modules built on the basis
of self-improving programs.

4. The indicated areas of application of artificial intelligence by criminals are obvi-

ous. They arise from the relation to the reality of those criteria for choosing a
field of activity, which are guided by organized crime around the world. But the
fourth area of application, at the first glance, belongs to science-fiction films and
novels. At least it has been repeatedly exploited there, and also formed the basis
of many cult portrayals, starting with the famous Terminator.
The idea of using a robot as a murder weapon is completely trivial. The first
person killed by a robot was Robert Williams, an American worker at the Ford
Motors Company killed at work on 25 January 1979. He worked as one of the
operators of the parts retrieval system. He climbed into the third level of the stor-
age rack, where he was struck from behind and crushed by one of the one-ton
transfer vehicles, killing him instantly. The robot was designed to retrieve cast-
ings from high-density storage shelves. Part of the machine included mechanical
arms to move items to and from the shelves. As a result of a glitch in the program

3
Digital Criminology

responsible for coordinating the arms, instead of a casting, it grabbed the worker
by the neck and strangled him. Since then, various undercover agents, journal-
ists, and scientists have repeatedly reported that crime syndicates pragmatically
and seriously discussed various options for murder, using electronics-rich cars,
smart homes, medical equipment, etc. In fact, it is quite possible to expect the
appearance of a fundamentally new phenomenon. The first known targeted as-
sassination by the means of an artificial intelligence system was the killing of the
Iranian General Qasem Soleimani in Baghdad airport on 3 January 2020 from
American drones.
Law enforcers around the world are seriously preparing for the emergence of un-
derground syndicates specializing in contract high-tech assassinations disguised
as technical incidents of various kinds. Given the volume of the contract killing
market in the United States of about $2 billion a year, the emergence of such
a high-tech criminal syndicate, and most likely not one, but several syndicates,
may be expected in the nearest future.
The main tool of such syndicates may not be hacker programs per se, but artifi-
cial intelligence. The subtlety here is the following: The vast majority of various
kinds of automated autonomous systems are controlled from a single comput-
ing centre that functions as artificial intelligence. This is called swarm training.
Accordingly, only another artificial intelligence can connect and replace the com-
mands of the artificial intelligence. It is beyond the power of a programmer; he
would be recognized due to the greater time lag, and less algorithmic actions
and operations.
In addition, only artificial intelligence is able to disguise a malicious shutdown
or execution of unauthorized actions as a technical failure. Despite some ex-
travagance, in the near future, this criminal business may become a reality. The
bad news is that, especially at the first stage, the overwhelming majority of such
murders will remain unsolved. There are simply no specialists in homicide de-
partments who can, on a professional level, understand the intricacies of neural
networks, deep learning, and active testing.

Criminals analyze events and processes and establish cause-and-effect relation-

ships no worse than university professors. Criminal organizations understand that
they cannot destroy or compromise law enforcement agencies’ databases. However,
the past need not be repeated in the future. If criminal organizations cannot destroy
law enforcement databases, then they will obviously go the other way. In any system,
humans are the most vulnerable link.
All over the world, law enforcement officers record attempts on the black market
to buy certain databases of images from video cameras installed in cafes, shopping
centres, police stations, government buildings, etc. This suggests that the criminals
have begun to create their own databases approximately utilizing the same artificial
intelligence solutions as law enforcement agencies. Given that the volume of their
databases will be significantly smaller, it is quite possible to implement it on open-
source artificial intelligence platforms by connecting them with commercially available
services for analyzing links, pictures, videos, texts, etc. Criminals will try to create

4
 Main areas of application of artificial intelligence by criminal groups

their databases by analyzing streaming videos from places close to law enforcement
buildings. First of all, it could be a database of undercover agents and informants.
Also, attempts to create databases on employees of police information centres, that is,
people admitted to the “holy of holies”, may be expected.

View publication stats