2019_2

American Journal of Research, Education and Development 2 ISSN 2471-9986 2019/4

American Journal of Research, Education and Development 3 ISSN 2471-9986 2019/4 CONTENT IoT architecture 1 Péter Török National University of Public Service, Hungary Scientific study of post-disaster recovery practice in Hungary, with special regard to private residential real estates COL József AMBRUSZ, Counselor University of Public Service, Disaster Management Institute Artificial Intelligence in cyberspace 1. AI-based cyber-attack capabilities András Tibor Fehér, Dr. Imre Négyesi National University of Public Service, Hungary

American Journal of Research, Education and Development 4 ISSN 2471-9986 2019/4 IoT architecture 1 Péter Török National University of Public Service, Hungary Abstract Nowadays, IoT is getting more and more a part of our lives. It also surrounds our homes, transportation, healthcare and industry. We can say that your everyday life has arrived. Therefore, it may be necessary to better understand its structure and how it works. In this publication, I want to help with that. I present a structural model made by various international organizations and companies and describe its architecture. Keywords: IoT, architecture, ITU-T, NIST, IEEE, ETSI, AWS, Azure, Intel SAS

American Journal of Research, Education and Development 5 ISSN 2471-9986 2019/4 Introducion Like the areas of application of IoT1, its structure is complex. Therefore, for perfect operation, its structure must be well thought out and its implementation post-standardized. When the average user hears the concept of IoT, he thinks of a smartwatch, a smart gadget at home, or a self-driving electric car equipped with internet communication. These thoughts are not far from reality. But IoT isn’t just about smart homes or self-driving cars. IoT has the potential to change everyday life and improve our quality of life at the individual level. At the societal level, bring about structural changes in its operation, in business, in the operation of the state, in public services. Ultimately, change the world. Solving such a lofty and large-scale task is a great challenge. One of the two arguably biggest challenges is the complexity of the IoT2, because of the variety of things you want to connect. The operation of any IoT system requires the utilization of all resources (hardware, software and systems), however diverse, to create a reliable and cost-effective solution integrated into a single framework. That is, every IoT deployment needs a rock-solid architecture to accomplish its intended purpose; the resulting efficiency and applicability of the system depends to a large extent on the quality of the infrastructure developed. IoT systems are different and have unique characteristics. But what they have in common is their architectural basis and their overall data flow process is pretty much the same. Every IoT system is based on connected devices, which are Internet-connected objects that can use their embedded sensors and actuators to detect the environment around them and gather information. These can be embedded in the devices themselves or connected as external peripherals. The next stage consists of data acquisition systems and gateways that collect large amounts of unprocessed data, convert it into digital data streams, filter, and preprocess it to be ready for analysis. It always operates in close proximity to the sensors and controls of specific devices. As an intermediary between networked devices, the cloud, and data analysis, gateways and data acquisition systems provide the necessary connection point that connects the other components. 1 Internet of Things 2 The other is the security.

American Journal of Research, Education and Development 6 ISSN 2471-9986 2019/4 The third stage is represented by edge tools that are responsible for further processing and enhanced analysis of the data. Visualization and machine learning technologies can also enter here. As the speed of data analysis is key in the application of some industrial IoTs, there has been a huge increase in interest in edge computing among industrial IoT systems recently. The data is then transmitted to cloud-based or locally deployed data centers. Here, data is stored, managed and analyzed in depth for actionable insight. A data center or a cloud-based system is designed to store, process, and analyze vast amounts of data to understand deeper relationships. Supported by powerful data analysis engines and machine learning mechanisms that edge systems could never support[1]. After a quick overview of the tasks to be performed by the system components, let's examine the details! Architecture models The biggest problem with examining IoT architecture is that it is not precisely defined. Depending on the source used, the IoT architecture can be two, three, four, or even five layers. Even the parts can change a lot. Therefore, there is currently no consensus on what the IoT architecture should look like. Compared to the TCP/IP3 model, which is the architecture of how the Internet works, there the components are very well defined. There is no doubt that Wi-Fi4 or Ethernet can only exist in the lower “Network Interface Layer” and HTTPS5 or IMAP6 can only exist in the upper “Application Layer”. These clear roles, distinct layers, are missing or have not yet developed in the IoT area. On the other hand, the lack of consensus in the IoT architecture is a reality that we have to deal with. The lack of a centralized architecture means that most IoT networks are developed on a case- by-case basis. There are currently two types of IoT architecture models: standardized and custom. In this case, standardized solutions, the competing international standards organizations, are their own solutions. The individual solution is an implementation specific to a manufacturer or group of manufacturers. 3 Transmission Control Protocol/Internet Protocol 4 Wireless Fidelity 5 Hypertext Transfer Protocol Secure 6 Internet Message Access Protocol

American Journal of Research, Education and Development 7 ISSN 2471-9986 2019/4 Security CapatiblesStandardized architecture models Management Capatibles Currently, several large standards organizations are working on their own architecture model. ITU-T Model: The Telecommunication Standardization Sector of the International Telecommunication Union (ITU-T) is now working on its own reference model. The draft is shown in Figure 1. The basic idea is that IoT devices can be categorized or classified according to certain similar properties. This allows you to define device classes with similar properties [2]. From this classification, architectural reference models can be designed that define the main functional entities of IoT devices and provide adequate funding for the implementation of solutions. Therefore, this recommendation defines the classification of IoT devices and presents a reference model of the architecture of each class IoT device. Application Layer Service & Support Layer Network Layer Device Layer Figure 1. ITU-T IoT reference model (Created by the author) In the model, the IoT devices are at the bottom of this along with the gateways. The networks are in the middle and the applications are at the top. Each layer has the capabilities of security and system management solutions [3].

American Journal of Research, Education and Development 8 ISSN 2471-9986 2019/4 NIST Smart Grid Model: For companies offering smart grids, the National Institute of Standards and Technology (NIST) has come up with an interconnected model that integrates service providers, markets, customers, distribution, transmission and generation into a single unit [4]. The model is shown in Figure 2. Figure 2. The NIST’s model [5] This was complemented by safety recommendations in the field of IoT. The subtitle of the recommendation is “A starting point for IoT device manufacturers”, the principles can be useful for anyone who connects a device to the Internet. This model specifies optional recommended cybersecurity features that should be built into network-capable devices, regardless of application. Core Baseline lists six recommended security features that manufacturers can incorporate into IoT devices: Device Identification: The IoT device must be able to identify itself. Device Configuration: Similarly, an authorized user must be able to change the software and firmware configuration of the device. Privacy: It should be clear how the IoT device protects the data it stores and sends over the network from unauthorized access and modification. Logical access to interfaces: The device must restrict access to local and network interfaces.

American Journal of Research, Education and Development 9 ISSN 2471-9986 2019/4 Software and firmware update: The device software and firmware must be updated using a secure and configurable mechanism. Cyber Security Event Logging: IoT devices should log cyber security events and make the logs available to the owner or manufacturer [6]. IEEE's7 three-tier architecture model: The IEEE considers the IoT architecture to be three- tiered, with all sensor devices treated as a base, network and data communication devices, and top- down applications. The best way to do this is to develop an IoT architecture that can be broken down into components. With a well-defined architecture, abstract ideas of the Internet of Things can be simplified and studied using concrete and known objects. In short, learning about each of the components of the IoT in an architecture can help expand our theoretical knowledge of the Internet of Things [7]. Applications Networking and Data Communications Sensing Figure 3. IoT-ARM (Created by the author) IoT-ARM8, developed by the IEEE, is an architectural reference model designed to connect vertically closed systems to create open systems as well as integrated environments and platforms. In this model, IoT stands for Connected Technologies, Devices, Objects, and Services. This reference model consists of several sub-models, of which the primary and mandatory model is the IoT domain model, which describes all concepts and their relationships that are relevant in the Internet of Things, such as devices, IoT services, and virtual entities. All other models, such as the IoT information 7 Institute of Electrical and Electronics Engineers 8 Internet of Things - Architecture Reference Model

American Journal of Research, Education and Development 10 ISSN 2471-9986 2019/4 model, functional model, communication model, security and data protection model, and IoT reference architecture, are based on the concepts introduced in the domain model [8]. ETSI architecture model for M2M9 communication: The European Telecommunications Standards Institute (ETSI) has developed a model for M2M devices that consists of two basic layers: a device and gateway domain and a network domain. Network Domain M2M Applications M2M Manegement M2M Service Capatibiles Network Core Network Manegement Access Network functions Device and Gateway Domain M2M M2M Applications Applications M2M M2M Service Service M2M Gateway M2M Device M2M Area Network M2M Device Figure 4. ETSI Communication Model for M2M Devices (Created by the author) This architecture consists of the following components: 9 Machine to Machine

American Journal of Research, Education and Development 11 ISSN 2471-9986 2019/4 • M2M-D10, a device running M2M applications that is usually embedded in a smart device and responds to requests or send data; • M2M area network, a capillary network (e.g., a small-scale home environment) created by a unique M2M-D using short-range communication technologies. • M2M-G11, the proxy responsible for the cooperation between the M2M local area network and the network domain. • Access Network: A network that provides access to the core network for devices (i.e., M2M- D and M2M-G) in the device gateway domain. • Core Network: A network that provides various services such as IP connectivity, network control, collaboration, roaming, and so on between M2M-A12 and M2M-D. It includes, but is not limited to, the 3GPP CN13, the ETSI TISPAN CN14, and the 3GPP2 CN.15 • M2M-A, M2M applications, application services that run service logic and apply M2M-SC16 through application programming interfaces. • M2M-SC, a network node that provides M2M functionality to M2M-A and hides network features for developing M2M applications [9]. Non-standardized architecture models. Despite the work of standardization bodies, many IoT companies have their own models, leading to separate IoT architectures. Amazon AWS17 IoT follows a three-tier IoT model that includes device software, connectivity and control services, and analytics services. AWS IoT Greengrass is a well-known example of device software [10]. The Azure IoT reference architecture is built upon Microsoft Azure platform to connect, store, analyze and operationalize device data to provide deep business insights. This architecture consists 10 Machine to Machine - Device 11 Machine to Machine - Gateway 12 Machine to Machine - Application 13 3rd Generation Partnership Project Core Network 14 European Telecommunications Standards Institute Telecoms & Internet converged Services & Protocols for Advanced Networks Core Network 15 3rd Generation Partnership Project 2 Core Network 16 Machine to Machine – Service Capatibles 17 Amazon Web Services

American Journal of Research, Education and Development 12 ISSN 2471-9986 2019/4 of core platforms services and application-level components to acilitate processing needs across three main areas of IoT solutions [11]. The Intel’s SAS18, whichis a reference architecture for IoT, i.e., for connecting products and services so that they can be aware of each other and surrounding systems in their ecosystems. There are two versions of reference architectures: version 1.0 for connecting the unconnected, using an IoT gateway to securely connect and manage legacy devices that are lack of intelligence and Internet connectivity; version 2.0 for smart and connected things, addressing security and integration capabilities that are essential for real-time and closed-loop control of the data shared between smart things and the cloud [12]. Conclusions By reviewing the IoT reference architectures, similarities in the development of several reference architectures can be observed. For example, the increase in the complexity and size of systems due to the huge amount of heterogeneous devices connected is the creation of domains. New collaboration solutions in the field of integrated information management. For greater efficiency and optimized processes or process chains. There is a growing need to achieve interoperability and compliance between different devices and systems. Finally, there is an increased demand for time to market and rapid development. Many of these drivers are also consistent with the goals described in the reference architectures. In surveying the reference architectures, I found that the business architecture and customer environment are often lacking. Most architectures, like the ITU-T and ETSI models, offer technical solutions, design patterns, and tactics. Among the reference architectures surveyed, the Amazon AWS-IoT architecture sample includes the edge layer, platform-level, and enterprise-level multi-level architecture sample, edge-to-cloud architecture sample, multi-level data storage architecture sample, distributed analysis architecture model, pattern, gateway or edge connection, and pattern of management architecture. However, models often lack business models and life cycle considerations. Yet reference architectures, like the NIST solution, should address technical architecture, business architecture, and the customer environment. Another important aspect of the reference architecture is to provide practice and guidance for creating new specific architectures. The best practices method is specifically suited for this task. IoT- 18 System Architecture Specification

American Journal of Research, Education and Development 13 ISSN 2471-9986 2019/4 ARM, which provides best practices and guidance for resolving specific architectures of IoT-ARM. It can also be used to develop system schedules that lead to minimal change between two product generations, thus guaranteeing system capabilities and features. The reference architectures presented are similar in their concept, but also differ in technology approach and implementation. Therefore, it would be difficult to draw a strict parallel along the definition of the IoT architecture and the TCP/IP model. Given the huge variety of IoT networks, we can best hope for the future. They will be compatible along some common elements or principles.

American Journal of Research, Education and Development 14 ISSN 2471-9986 2019/4 REFERENCES [1] BASSI, A. (Editor), BAUER, A. (Editor), FIEDLER, M. (Editor), KRAMP, T. (Editor), van KRANENBURG, R. (Editor), LANGE, S. (Editor), MEISSNER, S. (Editor): Enabling Things to Talk: Designing IoT solutions with the IoT Architectural Reference Model; Springer Heidelberg New York Dordrecht London, 2013. [2] ITU-T: Internet of Things (IoT); (2019.06.25.) https://www.itu.int/en/ITU-T/ssc/resources/Pages/topic-001.aspx; [3] Recommendation ITU-T Y.4460: Architectural reference models of devices for Internet of thingsapplications; (2019.06.25.) https://www.itu.int/rec/dologin_pub.asp?lang=e&id=T-REC-Y.4460-201906-I!!PDF-E&type=items; [4] NIST: Internet of Things (IoT); (2019.06.10.) https://www.nist.gov/internet-things-iot [5] WOLLMAN, D.: NIST Smart Grid Program –Overview for Smart Grid Task Force; (2019.06.10.) https://www.nist.gov/system/files/documents/smartgrid/NISTsmartgrid_overview_SGTF_wollman- Apr2011.pdf [6] NIST: Recommendations for IoT Device Manufacturers: Foundational Activities and Core Device Cybersecurity Capability Baseline; (2019.06.10.) https://nvlpubs.nist.gov/nistpubs/ir/2019/NIST.IR.8259-draft.pdf [7] IEEE: Internet of Things (IoT): Architecture and design; (2019.06.18.) https://ieeexplore.ieee.org/document/7759958 [8] IEEE: A General Architecture of IoT System; (2019.06.18.) https://ieeexplore.ieee.org/document/8005876 [9] Machine-to-Machine communications (M2M);Functional architecture; (2019.06.22.) https://www.etsi.org/deliver/etsi_ts/102600_102699/102690/02.01.01_60/ts_102690v020101p.pdf [10] AWS Architecture Center; (2019.06.04.) https://aws.amazon.com/architecture/ [11] Azure Application Architecture Guide; (2019.06.06.) https://docs.microsoft.com/en-gb/azure/architecture/guide/ [12] Architecture Specification White PaperInternet of Things (IoT); (2019.06.08.) http://d885pvmm0z6oe.cloudfront.net/hubs/intel_80616/assets/downloads/general/Architecture_Specific ation_Of_An_IOT_Platform.pdf

Scientific study of post-disaster recovery practice in Hungary, with special regard to private residential real estates1 COL József AMBRUSZ, Counselor 2 University of Public Service, Disaster Management Institute Abstract Since 2000 up to date, Europe has suffered over a hundred of flood phenomena. These floods demanded the lives of more than a thousand human lives, over half a million people had to be evacuated and temporarily resettled, the economic losses amount to at least EUR 30-50 billion. The extreme flood of the Danube in 2013, hitting Germany, Austria, Slovakia, Hungary and Croatia has further increased this number. Intense gales, torrential rainfalls, floods, inland waters and hailstorms, developed due to the extreme weather of the past decade, devastated in almost the entire territory of Hungary. Damages of different types and magnitude occurred in almost all the counties. They were of such severity that their rehabilitation could not be financed by the municipalities or the victims from their own resources. The governments of Hungary granted, sometimes voluntary, support and assistance to the municipalities and the local victims following the most violent and extensive natural disasters, and undertook the responsibility of centrally controlling their management. Keywords: Mitigation, elimination of consequences, disaster management, damage assessment 1 This study describes the standard procedures of rehabilitating private real estates damaged during disasters. 2 COL AMBRUSZ, József, Counselor, assistant professor; University of Public Service; Disaster Management Institute, H-1101 Budapest, Hungária krt. 9-11., [email protected]

American Journal of Research, Education and Development 16 ISSN 2471-9986 2019/4 Introduction Throughout history, many records were left behind for us, in which nature posed challenges to people against their own actions through its overwhelming power. Human beings, since their appearance, have been constantly striving to develop a safe environment, to satisfy their deficiency- based or growth-based needs. Mankind is seeking protection against hazardous natural processes and striving to avoid vulnerability. People tried, during the development process, with more and more complex technological solutions, to create, above all, a safe built environment. The conception of scientific circles is still divided whether today's extreme weather phenomena belong to the range of ordinary fluctuations, or there are noticeable signs of global climate change on Earth. Nevertheless, as the studies of a major project named “VAHAVA”, organized under the auspices of the Hungarian Academy Sciences, call the attention of organizations, having responsibility for preventing and responding to natural disasters and emergencies, to the fact that they should be prepared in due time for the occurrence of weather conditions that can nowadays be still regarded as extraordinary. Photo 1: 2001, Flood of the Tisza River, 2001. Photo by György Páros, NDGDM Climate change is a risk, threatening the national economy and calling for action, also for the Hungarian society. Based on versatile analyses, we can expect that temperature and precipitation conditions may significantly change in the coming decades, including a possible shift of the seasons, an increase of the intensity and frequency of particularly extreme weather phenomena that may jeopardize our values of nature, water resources, wildlife, forests, crop yields, buildings, residential environment,

American Journal of Research, Education and Development 17 ISSN 2471-9986 2019/4 public health and the quality of life as well. A UN group of scientists concluded that Hungary is one of the most vulnerable countries in Europe in terms of impact of the climate change on biological diversity, i.e. richness of species of the flora and fauna. Tens of thousands of family homes were destroyed or damaged throughout Hungary as a result of natural disasters, especially floods and inland waters, gales, heavy rainfalls, continuously recurring for over a decade. The creation of elementary conditions of housing – rehabilitation – was often beyond the possibilities of stakeholders. Neither the Government of Hungary, nor the European Union has statutory or contractual obligations to eliminate the consequences of natural disasters. Nevertheless, to mitigate the damages, the Government may decide to grant budgetary support as a voluntary commitment, based on the number of people affected, the extent of the damage, the economic, employment and social situation in the affected area. Following the most violent and extensive natural disasters the Government often granted, ad hoc voluntarily, support to the municipalities to assist local victims or undertook the responsibility of centrally controlling their management. The method of damage mitigation, the system of government support was regulated by individual government decisions before, and by individual government decrees in recent years. A specialized state-organized and state-controlled disaster management system, its method of management evolved slowly throughout history. At the same time, in relation to the synthesis of the triad of prevention, elimination and rehabilitation of the consequences of natural disasters, the impact of satisfying human security and safety needs upon several fields of science displays a change. Professional disaster management organs, continuously ever since their establishment, have been actively involved in the assessment and clarification of damages, have been present during damage estimations, in the operational staffs and the reconstruction committees. They have performed control tasks and monitoring activities, delivered assistance in the distribution of charity donations for those in need, analyzed information and forwarded the necessary survey data. The majority of municipalities accepted and required this professional involvement. Professional law enforcement organs expertly and accurately carried out the tasks related to the mitigation of damages, sometimes undertaking extreme loads, with the support of local residents. Hungary's primary natural, more specifically hydrological, threatening impacts are the damages originating from surplus waters.

American Journal of Research, Education and Development 18 ISSN 2471-9986 2019/4 The fundamental goal of rehabilitating the damages is to normalize everyday life and to achieve at least the condition before the disaster. Rehabilitation of damages, in municipal constructions, following the basic principle of municipality, is aimed at restoring the capacity to perform the required cooperation tasks again. The objective of rehabilitation of damages, occurred to private residential buildings, is to provide basic housing conditions for the victims in need. Taking into account the mitigation practices so far, in the case of ownership of individuals, in respect of the ownership, the main objective of the support is to provide basic housing conditions for victims using the buildings for residential purposes, nevertheless, analyzing the decision-making practice in mitigation, real estates in different ownership may also be granted support under specified or assumed conditions. However, non-residential buildings, residential buildings in different ownership, buildings used as rented apartments, rehabilitating rooms not necessary for normal housing and privately owned, but not habitually occupied properties shall not be subject to support either. In 2000, facing the challenges of sustained floods and inland waters, and in the emergencies (the Tisza River cyanide contamination, the fire in Ágasegyháza and the Pusztaszőlős gas eruption), the new disaster management organization was actively involved. Still, from the end of November 1999, inland waters emerged across the country and nearly 700 settlements conducted protection works. One of the most extensive and serious floods in the country's history was fought back in the spring of 2000, in which almost all of the Great Hungarian Plain was threatened by the flood. Damages were caused to nearly thirty thousand residential buildings in personal property and two thousand buildings performing municipal duties. In March 2001, the upper Tisza River flood inundated a part of the Bereg region, lying between the state border and the Tisza River. Huge losses were recorded, roads, bridges, ferries, rail and flood control facilities were damaged, residential buildings and buildings essential for the people's and their communities' lives were destroyed or damaged. The Government of Hungary made a commitment of HUF 22.8 billion to the necessary rehabilitation and reconstruction works, furthermore, established a Rehabilitation and Reconstruction Committee led by the Director General of NDGDM, for the coordination of mitigation. There had been no instances of performing such complex tasks seen since 1970, the centralized organization of managing the mitigation was motivated by the concentrated characteristics of the damages, the large number of elderly victims and the need of preservation of the characteristics of folk architecture.

American Journal of Research, Education and Development 19 ISSN 2471-9986 2019/4 In the period between 9 and 12 August 2002, in Borsod-Abaúj-Zemplén, Heves and Nógrád Counties, due to torrential rainfalls, in 43 settlements, 268 residential houses and 54 buildings performing municipal tasks were damaged. From 7 to 9 August 2002, as a result of extremely large amounts of rainfall, extreme floods rolled down along the Danube and its tributaries. The high water level caused damages in 8 counties, in 674 residential properties, and nearly 80 municipality-owned properties. In relation to damage mitigation, the Government initiated the formation of county committees of rehabilitation and reconstruction, as well as the financial support to be delivered to the victims. In June 2004, heavy rainfalls, accompanied by strong winds and hail, caused severe damages in Bács-Kiskun, Békés, Borsod-Abaúj-Zemplén, Komárom-Esztergom, Nógrád and Tolna Counties and the capital, Budapest. In order to help rehabilitate the residential properties, owned by individuals, damaged due to the extreme weather, the victims received financial support with the involvement of municipalities; for this purpose the government allocated HUF 400 million. In 30 municipalities, nearly 1,900 property owners received support within the framework of financial aid granted by law. At the end of July, due to the large amount of rain in Slovakia in the catchment area of the rivers of Borsod-Abaúj-Zemplén County, a flood occurred along both Hungarian and foreign watercourses, as well as on the Hernád River. Despite all efforts, there were settlements, or parts of them, which could not be protected from the flood, so they were all inundated. The Government also tried to assist the owners of residential properties, damaged by the flood: an amount of HUF 50 million was granted to mitigate the damages to individual owners of residential properties. 27 municipalities were affected by the flood, in 16 settlements 207 buildings were damaged, three of them collapsed. The proceedings, initiated in connection with the gale and flood damage, were completed the following year and the secure housing conditions of the victims and their families were resolved. 2005 After very serious natural disasters, the Government granted support to private property owners of residential buildings on six occasions: Pásztó-Mátrakeresztes, Mád, Boldva, Várvölgy, Hajdú-Bihar, Szabolcs-Szatmár-Bereg, Borsod-Abaúj-Zemplén, Jász-Nagykun-Szolnok, Zala County. According to a survey carried out with the participation of the disaster management organization, a total of 5,526 residential properties were damaged in a total of 147 settlements due to

American Journal of Research, Education and Development 20 ISSN 2471-9986 2019/4 various natural disasters. For damage mitigation, the Government allocated HUF 1,430.6 million. The Ministry of the Interior raised the amount to HUF 1,661.0 million. 2006 In order to mitigate the damages caused by the spring flash floods along Hungary rivers, and the inland waters during the first months of the year in certain territories of the country, the State helped the municipalities like in the previous years, to be able to support their residents in need. The Government granted HUF 3.1 billion to mitigate the damages of the 10,328 residential properties. An additional support in 10 counties and the capital city involved nearly 1,450 damaged properties. 2008 - 2009 As in the previous years, in 2008 as well, significant damages occurred both to municipal and private buildings due to extreme weather conditions - gales, hail, torrential rainfalls. NDGDM was involved in organizing the damage assessment, systemizing the damage data and preparing decisions. 7 June 2009, in Szabolcs-Szatmár-Bereg County, along the line Tiszavasvári- Vásárosnamény-Beregsurány, a hail with violent storms moved through an approx. 10-km wide strip, causing considerable damage to 7,853 residential properties and 180 municipal and municipally maintained properties. From 15 May 2010, high-intensity rainstorms affected almost the entire country from a cyclonic front, accompanied by stormy gusts (sometimes exceeding 90 km/hour). Because of the sudden and significant amount and intense rainfalls, creeks (Boldva, Szinva, etc.) and rivers (Ipoly, Hernád, Sajó, Bodrog, Zagyva, etc.) inundated in a number of places, not sparing the local government buildings or residential houses either. In total, 49 municipalities in 8 counties, 271 property owners received state financial aid. In 81 cases, victims chose the construction of new properties, in 190 cases, second-hand houses were purchased for compensation. 1 billion 780 million 585 thousand Ft grant allocations were paid as central compensation. Charitable services provided additional funding to reconstruct properties. In order to establish appropriate flood protection of settlements damaged during the flood in May and June 2010, and reconstructed afterwards, the Government declared emergency in Borsod- Abaúj-Zemplén County. In order to avoid the recurrence of the devastation of 2010, the Government decided on investments for flood damage prevention in its Decree 1028/2011. The decision was designed to initiate flood protection investments in Borsod-Abaúj-Zemplén County that would serve the protection of the life of the population and property. Flood protection investments accomplished and handed over:

American Journal of Research, Education and Development 21 ISSN 2471-9986 2019/4 • 4 rain shower reservoirs in Hídvégardó, Bódvaszilas, Szendrőlád and Edelény, • 3 emergency reservoirs in Bódvalenke, Bódvarákó, Bódvaszilas, along the Bódva Creek, • Felsőzsolca is protected by circular levees built with different technologies, • Ónod and Nagycsécs are protected by circular levees built by the municipalities, • Riverbed dredging was carried out on the following creeks and rivers: Sajó, Bódva and Hernád The Government also decided on the elaboration of the water damage recovery plans of the 98 most endangered settlements, which is one of the most effective elements of flood control and response. All municipalities received this plan. The primary task of the four rain shower reservoirs is to retain the rain suddenly fallen in a given small area. The three emergency reservoirs have the function in the flood peak period: at a given moment, they are opened to drain the excess water to relieve the load on the responders and protection structures in the lower section of the river.3 The floods of 2013 in Central Europe, at the end of May and early June, occurred mainly in the areas of Austria, Czech Republic, Hungary, Germany, Switzerland and Slovakia, originating in the Danube and Elbe Rivers and their catchment areas. In Germany, 178% of the precipitation of average of many years fell in May, and the intense rainfalls there and in Austria greatly contributed to the development of emergency. Austria Due to the inundated rivers, many roads were closed; traffic was clogged. Railway lines were inundated at several places. From noon 3 June, there was a shipping ban on the entire stretch of the Danube in Austria. Six persons died. One was missing. Vienna was not endangered; last time it happened was in the 1970s. Czech Republic The Czech government declared emergency on 2 June because of the weather and the flooding of the Vltava and Elbe Rivers. In Prague, there were creeks which would do no harm in normal times, but due to the flesh flood, they caused immense trouble. 11 persons died, four were missing, approx. 700 settlements were stricken by the floods. 20,000 persons had to be evacuated. 3 Source: NDGDM

American Journal of Research, Education and Development 22 ISSN 2471-9986 2019/4 Germany The most severe situation was at the meeting points of three rivers in Passau: Danube, Inn and Ilz. The water level of the Danube exceeded the one measured in 2002: 10.8 meters, then topped at 12.89 meters. Slovakia On the entire stretch of the Danube in Slovakia, 3rd degree preparedness was ordered in five districts. The situation was the most critical at Dévény. In Bratislava, the river topped at 10.32 meters. Hungary It is important to note that all flood waves are different and there are no two identical. Meteorological and hydrological factors play a joint role in their development. This flood wave was caused by a significant amount of precipitation of four consecutive days. It resulted in record high water levels in most parts of the Danube’s stretch in Hungary, so new “highest water level” (HWL) values were reached.4 Settlement New HWL[cm] Old HWL [cm] Time of old HWL Nagybajcs 907 872 2002 Komárom 845 802 2002 Esztergom 813 771 2002 Nagymaros 751 714 2006 Budapest 891 860 2006 Dunaújváros 755 742 1965 Dunaföldvár 721 703 1965 Paks 891 872 1965 Dombori 916 894 1965 Baja 989 976 1965 Mohács 964 984 1965 4 Source: National Meteorological Service

American Journal of Research, Education and Development 23 ISSN 2471-9986 2019/4 Methods In this paper, a secondary research was performed by the author. During this secondary research, literature review and report content analysis I was implemented with a focus on the characteristics of post-disaster rehabilitations and benchlearning opportunities connected to Hungary’s recent floods. Several searches I was run on Google, Research Gate, Elsevier ScienceDirect and EM-DAT (Emergency Events Database) databases, with the following queries: “rehabilitation in Hungary”, “reconstruction after the flood in Hungary”, “post-disaster reconstruction”, “compensation”, “reconstruction in Hungary”, “disaster management”, “civil protection”, “benchlearning” and their different synonyms. In the following, relevant papers, reports, datasets I was scrutinized from the search results and was used for this paper. 1. LEVELS OF THE ORGANIZATIONAL SYSTEM OF DAMAGE MITIGATION As far as the levels of controlling damage mitigation operations, they may be differentiated by the types of support to be implemented through central organization, from the funds available at municipalities and, last but not least, through the municipalities. The selection of the type of control can basically depend on the damages to life and property safety and also on its extent, on the type, intensity of the incident, on the size of the area damaged, on the number of victims, on the proportion of assets and personnel available, and also depends on whether governmental actions are needed or not in order to eliminate the circumstances (rehabilitations). Appointment of Government Commissioner, temporary Government Commissioner; establishment of an inter-ministerial committee The Government, by issuing a decree, may appoint a commissioner to perform extremely important tasks or in case an issue does not belong to the scope of any ministry. On this person, the minister responsible for the coordination of the work of the Government makes proposal. The appointed is an entity with a minister’s rank, receiving authority to coordinate a certain range of issues, perhaps to solve, manage or control a specific issue.[1] In order to prepare legislation and to coordinate inter-ministerial affairs successfully, an Inter- ministerial Committee for the Coordination of Reconstruction and Rehabilitation may become necessary depending the severity and extent of the incident.

American Journal of Research, Education and Development 24 ISSN 2471-9986 2019/4 2. NOTIFICATION OF DAMAGES The assessment and estimation of damages may start on a sole governmental decision concerning the deadline of notification. The damage assessment is preceded by the notification by the victim. The mayor is responsible for organizing the notification of damages. The mayor announces the method and deadline of the notification of damages as locally usual. The applicant makes a declaration on a form whether he consents to the management of his data, to the transfer to a third party associated with the mitigation. The notary must keep a record of the damages reported, and he prepares a summary on it within 5 working days after the deadline for the notification of damages. This is sent to the competent county disaster management directorate with a copy of the notification forms each. He may also initiate to the Chairman of the Protection Committee to appoint a local reconstruction commissioner to coordinate the damage assessment team(s), as well as to perform additional rehabilitation and reconstruction tasks. During professional verification, it is necessary and appropriate to compare the reports with the already available damage data and areas affected by the disaster. 3. DAMAGE ASSESSMENT METHOD After a natural disaster, the damage assessment is performed by the capital’s and county government offices, geographically competent, by the county disaster management directorates, and, if applicable, pursuant to the request of the capital’s and the county government offices, geographically competent, in the case of protection against floods or inland waters, by the Environment and Water Directors, geographically competent, in the case of damages to roads, ferries and bridges, by a representative of the Coordination Center for Transport Development, and in the case of collapse of cellars or embankments or landslides, by a representative of the Commission of Experts on Cellar and Embankment Emergency Response.[2] If the damages in a settlement or in a part thereof affect a minority, by a significant proportion, and the minority has a minority self-government in the locality, the minority self-government is to be asked to participate in the work of the damage assessment teams. Prior to beginning of damage assessment activities, the Chairman of the capital’s or county Protection Committee, in cooperation with the territorial bodies of the professional civil protection agency, organizes a training for the members of the working group on damage assessment guidelines and methods, key construction standards and local specificities. The members should be provided with credentials, signed by the county or the capital’s Chairman of the Protection Committee. Damage assessment forms must be

American Journal of Research, Education and Development 25 ISSN 2471-9986 2019/4 filled out in triplicates. Damages must be documented by photos taken during damage assessment. The contents of the damage assessment form is may be supported by other documents (property registry, contract, layout design, previous valuation, expert, witness, protocol, photograph, video, declaration on being insured, etc.). A copy of the damage assessment form remains at the office of the mayor; for municipalities operating district notary offices, at the district notary office. The mayor shall forward a copy to the disaster management directorate and to the victim.[3] Depending on the nature of damages, additional tasks must be determined, resulting from the damages occurred (soil mechanics analyses, the scope of secondary, collateral damages, static tests). The valuation is the amount calculated based on m2/forint price provided by the tax agency with territorial jurisdiction. The assessment of damages can be monitored and controlled by the central organ of the professional body for disaster management. The county disaster management directorate summarizes the damage data and the rehabilitation, reconstruction, purchase cost estimation data received form the settlements of the county, and submits them through the central organ of the professional disaster management body to the minister responsible for the protection against disasters. Against this background, the Government decides on the nature and extent of compensation and provides for the possibility of reimbursing the costs of damage assessment. The minister, based on the Government's decision and the damage assessment data, determines the amount of each settlement’s support allocation. 4. A POSSIBLE WAY OF CALCULATING THE AMOUNT OF FINANCIAL ASSISTANCE The determination of support of rehabilitation in the case of individuals assumes the benchmark that quantifies the degree of support on the one hand; and compares the level of support with the damage value and with the amount paid by the insurer, as well as takes into account the victim's social status and the numerical value of his own resources.[4] In relation to the previous rehabilitation, several support calculation methods are available as well. One way is to determine the percentage of local allocations, such as in the case of the compensations in 1999 or 2006, damage values resulting in individual victims’ properties or municipal allocation figures, bound to other conditions, which may be modified by a social decree of

American Journal of Research, Education and Development 26 ISSN 2471-9986 2019/4 a municipality (e.g., not having insurance or WMA (*) policy, in the case of 50% of the assessed cost of rehabilitation of residential properties).5 Another method of calculation is the amount of damage value projected to the square meter damaged, in which the damage value and the actual rehabilitation value may be separated. Considering the conditions laid down: Amount of damage mitigation = Costs of actual rehabilitation - (Amount reimbursed by insurance + own resources of victim) The coefficients of the above formula are not constant. The application of the formula is made more difficult by the fact that the support of owners of residential properties in private property can be implemented by granting social and housing support forms as well, based on local decrees of municipalities. The amount resulting from home insurance must be taken into account, based on self- reporting of victims. The calculation is even more difficult because of peculiarities of construction implementations, in many cases, of unsettled ownership, any differences between statuses registered in the land registry and the real ones.[5] 5. CONCLUDING SUPPORT AGREEMENTS The compensation support may be paid based on the damage notification by the victim and its adjudgement, as well as on the grant agreement. The agreement, taking into account the considerations of need, as determined by the act on social administration and social benefits, is concluded by the local government. It should include the type and amount of compensation, the amount of support, payment dates on behalf of the local government, adjusted to the construction; the stipulation of control entitlement of first and second instance building authority and regional and local organs of the professional disaster management organization, the purpose of use and documents of proof of proper use, accountability, its method and the final deadline, and mortgage and the provisions for the ban on sale and debit; clause on the repayment liabilities in the event of use other than the intended purpose, and the victim's statement of the existence of insurance contractual relationship or undertaking of the obligation of contract. It should include the rules of purchase of second-hand 5Wesselényi Miklós Flood and Inland Water Protection Compensation Fund

American Journal of Research, Education and Development 27 ISSN 2471-9986 2019/4 housing, or in the case of new construction of housing, of ownership of the local government and the rules for accounting. The victim must prove to the notary the conclusion of an insurance contract, and the fact that insurance policy may not be concluded with an insurance company, latest at the conclusion of the agreement.[6] 6. OTHER FEATURES OF CONSTRUCTION The relevant legislation displays it markedly that the reconstruction of destroyed buildings shall not be allowed in a place where there is an increased risk of the possibility of the occurrence of natural or man-made disasters. Reconstruction can only be done in an area designated by a resettlement plan or in a pre-approved area. By application of exceptional building regulations related to the rehabilitation of building damages occurred due an emergency declared, represents a change in the built environment and a ground for prevention. The notary or an administrator of the office of the mayor commissioned by it, as well as the regional and local organs of the professional disaster management authority controls the construction on the site.[7],[8] Photo 2: Reconstructed residential buildings, 2001, Bereg Photo by György Páros, NDGDM SUMMARY The general legal framework of rehabilitation and reconstruction of private residential properties developed a few years ago. Through the adoption of rules relating to recovery and reconstruction, included in the Chapter XI of Government Decree 234/2011, on the implementation of Act CXXVIII of 2011 on disaster management and the amendment of certain acts in relation, through the development of procedures for decision, created a regulated management plan; for the

American Journal of Research, Education and Development 28 ISSN 2471-9986 2019/4 sake of responsible, cost-effective and legal use of public funds, the reduction of the extent of the damages to residential property after natural disasters, in order to promote the mitigation of damages, provided a possibility to use various forms of self-care (being insured, permanent maintenance of buildings and the residential environment). However, the rehabilitation of buildings of local governments is carried out in a specific regulatory framework, which support includes specific procedural rules. Lessons learnt from rehabilitation and damage incidents are necessary to be re-implemented partly in prevention, within the framework of authoritative specialized authoritative procedures, enforcing regulations, on the other hand, in the harmonious unity of developments and investments, and for the reverse resolution of force majeure conditions, which cannot do without international cooperation. European regional cooperation should be continuous; therefore, considering the responses to the challenges of disasters of natural origin, joint action programs should be implemented, creating the system of prevention, built on each other, the community interests of damage mitigation.

American Journal of Research, Education and Development 29 ISSN 2471-9986 2019/4 REFERENCES [1] Complex Kiadó: A kártérítési jog magyarázata, comment, ISSN 1589-0058 (Professional editor: Fézer) [2] Bukovics, István PhD: Katasztrófaigazgatás, ISBN 978-963-9313-77-4 [3] Kozári, László PhD – Simon, Ildikó: A globális klímaváltozással összefüggő katasztrófavédelmi zabályozási, vezetésirányítási, szervezési kérdések vizsgálata http://www.vedelem.hu/letoltes/tanulmany/tan170.pdf [4] Restás, Ágoston: Az UAV katonai alkalmazásának transzfere a polgári alkalmazás felé: Katasztrófavédelmi alkalmazások, REPÜLÉSTUDOMÁNYI KÖZLEMÉNYEK 25 (2), pp. 626-635 Decision-making Mechanism of Firefighting Managers; Védelem, year VIII, No. 2, Budapest, 2001, ISSN: 1218-2958 [5] Janos Bleszity, Lajos Katai-Urban, Zoltan Grosz: Disaster Management in Higher Education in Hungary, ADMINISTRATIVA UN KRIMINALA JUSTICIJA - LATVIJAS POLICIJAS AKADEMIJAS TEORETISKI PRAKTISKS ZURNALS 67: (2) pp. 66-70. [6] Muhoray Árpád: A katasztrófavédelem aktuális feladatai Hadtudomány (online) 2012. 1-16 o. Forrás: http://mhtt.eu/2012/2012_elektronikus/2012_e_Muhoray_Arpad.pdf (Letöltés ideje: 2019.07.31.) [7] Gyula, Vass ; János, Bleszity: The Analysis Of The Process Of Rehabilitation Following Disasters, VÉDELEM TUDOMÁNY : KATASZTRÓFAVÉDELMI ONLINE TUDOMÁNYOS FOLYÓIRAT V:1 pp. 98-117., 20 p. [8] Dobor, József ; Kuk, Enikő ; Kóródi, Gyula ; Kocsis, Zoltán: Industrial Safety Analysis of Accidents Involving Ammonia, with Special Regard to Cold-Storage Facilities II, ACADEMIC AND APPLIED RESEARCH IN MILITARY AND PUBLIC MANAGEMENT SCIENCE 15 : 1 pp. 37-49. Paper: 2064-0021, 13 p. (2016) Downloads: http://www.slidefinder.net/a/ambrusz/34ambrusz/30413329 www.vedelem.hu/files/UserFiles/File/konf2010/.../34.ambrusz.ppt http://www.katasztrofavedelem.hu/letoltes/magazin/k200607.pdf?rand=118 http://tuzoltoborze.hu/index.php?option=com_content&view=article&id=626%3Aaz-arvizkarok-elzetes- ertekbecslese&Itemid=185 Legislation (Acts -, Government Decrees): 1. Magyarország Alaptörvénye 2. 2011. évi CXXVIII. törvény a katasztrófavédelemről és a hozzá kapcsolódó egyes jogszabályok módosításáról, 3. 234/2011. (11. 10.) kormányrendelet a katasztrófavédelemről és a hozzá kapcsolódó egyes jogszabályok módosításáról szóló 2011. évi CXXVIII. törvény végrehajtásáról, 4. 2011. CXIII. törvény a honvédelemről és a Magyar Honvédségről, valamint a különleges jogrendben bevezethető intézkedésekről, 5. 290/2011. (XII. 22.) kormányrendelet a honvédelemről és a Magyar Honvédségről, valamint a különleges jogrendben bevezethető intézkedésekről szóló 2011. évi CXIII. törvény végrehajtásáról. 6. a Magyar Köztársaság 2011. évi költségvetéséről szóló 2010. évi CLXIX. törvény 7. a számvitelről szóló 2000. évi C. törvény 8. az általános forgalmi adóról szóló 2007. évi CXXVII. törvény 9. a közbeszerzésekről szóló 2011. évi CVIII. törvény 10.az államháztartás működési rendjéről szóló 292/2009. (XII. 19.) Korm. rendelet 11. a társasági adóról és az osztalékadóról szóló 1996. évi LXXXI. törvény 12. a szociális igazgatásról és szociális ellátásokról szóló 1993. évi III. törvény 13. a kedvezményezett térségek besorolásáról szóló 311/2007. (XI. 17.) Korm. rendelet 14. a társadalmi-gazdasági és infrastrukturális szempontból elmaradott, illetve az országos átlagot jelentősen meghaladó munkanélküliséggel sújtott települések jegyzékéről szóló 240/2006. (XI.30.) Korm. rendelet 15. Magyarország helyi önkormányzatairól szóló 2011. évi CLXXXIX. törvény 16. a vizek kártételei elleni védekezés szabályairól szóló 232/1996. (XII. 26.) Korm. rendelet 17. a vis maior tartalék felhasználásának részletes szabályairól szóló 9/2011. (II. 15.) Korm. rendelet és 2013. évi módosítása

American Journal of Research, Education and Development 30 ISSN 2471-9986 2019/4 Government Decrees (Korm. h) and Government Decisions (Korm. r.) used occasionally: 1042/1999. (IV.29.) Korm. h; 1091/1999. (VIII.13.) Korm. h; 1092/1999. (VIII.13.) Korm. h; 1058/2000. (VII.11.) Korm. h; 1019/2001. (III.9.) Korm. h; 1025/2001. (III.23.) Korm. h; 1/2001. (III.26.) HUB h; 1033/2001. (IV.12.) Korm h; 3/2001. (V.7.) HUB h; 1104/2001. (IX.12.) Korm h; 1054/2003. (VI.13.) Korm h; 1142/2002. (VII.16.) Korm h; 1155/2002. (IX.14.) Korm h; 1148/2002 (IX.5.) Korm h; 1149/2002 (IX.5.) Korm h; 201/2002. (IX.14.) Korm r; 1180/2002. (X.24.) Korm h; 217/2004. (VII.19.) Korm r; 1072/2004. (VII.19.) Korm h; 243/2004. (VIII.21.) Korm r; 259/2004. (IX.16.) Korm r; 1090/2004. (IX.16.) Korm h; 88/2005. (V.25.) Korm r; 93/2005. (V.21.) Korm r; 1050/2005. (V.21.) Korm h; 99/2005. (V.28.) Korm r; 1054/2005. (V.28.) Korm h; 151/2005. (VIII.2.) Korm r; 158/2005. (VIII.16.) Korm r; 181/2005. (IX.9.) Korm r; 1088/2005. (IX.9.) Korm r; 155/2006. (VII.26.) Korm r; 2132/2006. (VII.26.) Korm h; 241/2006. (XI.30.) Korm r; 2078/2007. (V.9.) Korm h; 221/2007. (VIII.23.) Korm r; 2156/2007. (VIII.23.) Korm h; 1148/2010 (VII. 8.) Korm. h; 1792/2013. (XI. 7.) Korm. h.

Artificial Intelligence in cyberspace 1. AI-based cyber-attack capabilities András Tibor Fehér, Dr. Imre Négyesi National University of Public Service, Hungary Abstract: Artificial intelligence (AI) has many uses. Below, we review the area of cyber-attacks in general. We will look at specific examples of attack models in the second part of this article. Reviewing the attacks, we show that since AI has become unavoidable, the realistic and practical way of applying the technology in the Hungarian Defense Forces must be constantly rethought. Keywords:

American Journal of Research, Education and Development 32 ISSN 2471-9986 2019/4 Introduction Artificial intelligence has spread around the world built into various technologies. Meanwhile, the word “smart” has become so fashionable that it is gradually being emptied as it is already associated with all kinds of products. By the term “teachable machines”, the journalistic or advertising language often means a classical program based on a database, which stores e.g. personal habits and later takes account these. Moreover, it often just means that a traditional device (clock, pen, fake nail, etc.) is augmented with some novel features, including a microchip. Although the fridge is not yet \"smart\" to order from the Internet what is running out: the program for this can be written by a skillful sixth-grader1. In short: “smart” machines may not be “intelligent”. The other extreme is represented in the public consciousness by the cyborgs of sci-fi's, who are completely human-like, have emotions, and even consciousness, or even religious faith, and perhaps represent the next step in evolution. In the following, we will NOT use the term AI in these colloquial terms. The concept of artificial intelligence can also be interpreted in a broad spectrum within scientific use. More expensive and less expensive, hardware and software AI solutions are available – just as an electric locomotive’s electric motor and a winchester positioning motor operate on the same principle, yet they are completely different in design. Many articles and books deal with neural network, machine learning, deep learning, and other related concepts, or AI divisions. These models and definitions are not detailed now so that we can focus on the subject. In the methods discussed, we examine techniques that operate with such “real” AI models, going far beyond the vernacular term “smart”. Many articles and books deal with a neural network, machine learning, in-depth learning, and other ideal terminology, or various divisions of AI [1]. In the methods discussed below, we examine techniques that operate with such “real” AI models, going far beyond what is associated with the colloquial term “smart machine”. But before we enter the cyber world, it is important to get around the more general significance of AI technology, as AI is slowly in its 60s, and yet we are not served by humanoid robots in our homes. 1 Some programmable games have been recommended by their manufacturer since the age of 7, with a programming interface that can be used without reading knowledge

American Journal of Research, Education and Development 33 ISSN 2471-9986 2019/4 How significant is AI in the world and in cyberspace? It is only recently that the era has come in which artificial intelligence is becoming a key part of our world [2]. There are those who think that it will transform the world like an explosive engine, but in our opinion, it would be more accurate to put it in parallel with nuclear energy. Quantum physics has also been part of the debates in the scientific world for a long time, and using only many other discoveries, it has become a matter of practice over several generations (we have energy from it, the basis of many cures, etc.). We still don't see his prospects exactly today. In this way, AI will be able to change our everyday lives with several generations after its mathematical foundation today. Most people don’t understand any of them, but everyone enjoys their results. In addition, the parallels between AI and nuclear science also point well to the dangers. It is very fortunate that we do not have the opportunity to learn the power of AI in the form of military developments, as was the case with nuclear energy. But unfortunately, there may come a time like the fear of the atomic bomb apocalypse in the Cold War era. In art, the dystopia of robotic apocalypse has already taken over the emphasis. The arms race has not yet slipped out - fortunately only in movies have such technologies reached terrorists. But in addition to the parallel, there are also many differences between nuclear and AI technology. From a protection point of view, the most significant is that a nuclear charge and the controls required for it can be physically better protected than a dangerous artificial intelligence code that can be stored on a pen-drive, and can be used on commercially available devices. It is to be hoped that the defence will be the stronger, and that the AI will be for the benefit of humani-ty, rather than be a more dangerous destructive force than an atomic bomb, as many renowned scientists and researchers warnes us [3]. For example, the issue of banning the Lethal Autonomous Weapon System (LAWS), which the UN has been addressing since 2013 [4], is of concern to many, but the leading superpowers are not yet inclined to do so [5]. Although the EU and Hungary are in favor of a ban [6], and several giants companies have declared that they will not participate in such developments, the covert development of LAWS systems has not stopped. There are countless other military uses for artificial intelligence, from support for planning tasks to military training. Yet the cyber battlefield is what is most endangered at the moment. This is because in other areas of the military industry we are still in the phase of prototyping, research and testing [1], but a new generation of attack and defense methods has been working in cyberspace for years, operates by AI. In this study, we only review the published part of the topic, as one of the

American Journal of Research, Education and Development 34 ISSN 2471-9986 2019/4 methods of “using AI in cyberspace” also fills thick books. Behind these can conjecture additional, non-public results in both defensive and offensive solutions. Methodology At the beginning of the research, it may be a problem how to study the military use of AI in cyberspace, i.e. the methodology itself is questionable. There are many things that make the investigation difficult, we can highlight two of them: 1. In cyberspace, all borders are blurred, not just national borders: in today's complex warfare, military attacks do not necessarily target the enemy's military side, but their business, state, cultural side, social media, and so on. The cyber battlefield wants to turn welfare technologies against man, and it has taken every opportunity to do so - so it will also use the results of AI's peaceful developments. For these reasons, we want to focus our research in vain on the military use of AI in cyberspace, because it is inseparable from civilian technologies. 2. Another typical barrier to researching military aspects is data sensitivity. For this reason, it is not possible to obtain ready-made research materials from various forces, but even if we get insight, these results remain secret and cannot be published. So we will not get a sufficiently thorough picture of combat cyber methods using AI from military sources2. It is assumed that the offensive technologies that can be applied in the cyber battlefield in the case of military industrial developments are not lagging behind civilian or underworld developments, so they either move forward or go head-to-head3. Therefore, it can be a perfectly legitimate method to examine what can be examined well: the research and business spheres. In this way we can make the current development directions and effects visible. From this it can be deduced the development of state or military threats. By examining some working attack models and hypotheses, possible directions will also be outlined. There is no datas that AI-based cyberspace is a pull sector in military development. The various military industries focus on the use of AI in other military areas. Nevertheless, it is certain that secret developments of this kind are underway, but there is no indication that any military research would significantly precede the use of non-secret cyber-AI. 2 There are analyzes of the status of military use of AI, but it is hardly to be found specifically in the cyber area. The role of AI in NATO will be examined in a separate study. 3 There is no evidence that military development is a leader industry in the cyber space, as has been the case for many technologies throughout history.

American Journal of Research, Education and Development 35 ISSN 2471-9986 2019/4 So below, we start with the corporate application of the use of AI for cyber purposes and look at scientific research. The technological lessons thus obtained are examined in the light of available, current and published political and military documents. From all this, conclusions can be drawn about the expected directions of the development of technology for military purposes. Lessons from the evolution of attack & defense to date More and more vulnerabilities are appearing in computing, as attackers try to perform malicious operations with new abilities. To this day the defense innovations have found any antidote to these. To these days, he has managed to defend about as much as the police of a well-organized, rich democratic state can keep crime at bay. Attack methods always include the previous ones, supplemented with new ideas, so it is worth reviewing the basic layers of today's attacks. Early computing and the beginning of the PC era were overshadowed by virus news, and everyone rightly feared the damage they caused. Many programming companies therefore specialize in this, and have responded with various antivirus software that can easily find malware based on a database. At the beginning of the web era, viruses spread too fast, that challenge was addressed by the heuristic search engines, and the active involvement of customers. Another challenging IT security task at the time was cross-site request forgery (CSRF) and cross-site scripting (XSS), for which firewalls provided a solution. Security has become an industry and complex security solutions have emerged that have provided protection – for a time. Then in the age of clouds, “smart” devices, and social networking: came the misuse of individual data, and the advanced persistent threats (APTs) – we do not yet have sure armor to defend them, while we have entered a new era, the era of AI. The transition to machine learning and AI has already begun about twenty years ago, but it is only in these years that it becomes the main direction of information technology – and with it, the main direction of cyber attacks. Experts agree that we are facing a new generation of threats that will also rewrite the protection principles we have been accustomed to so far. According to Mikel Rodriguez (MITRE Institute for Machine Image Recognition Research), due to AI, the situation of cybersecurity has now become similar to the challenges of the mid-1990s [7]. So it is not enough to hone the defense, as we did for fifteen years, but must to put it on a whole new footing. The question is who can make more effective use of the learning and decision-making capabilities of computers: cybercriminals and military cyber-attack systems - or cyber-defense. Approaching the answer, it is not expedient to talk about AI in general, as it is an umbrella term, that includes many types of algorithms. Some of these can only be specified for attack or defense only,

American Journal of Research, Education and Development 36 ISSN 2471-9986 2019/4 and many can be specified for both purposes. To review the situation, we first examine how cyber- attacks are expected to change in a changed technological environment if the potential of AI is to be exploited. The main possibilities of AI attacking in cyberspace To review the situation, we first examine how cyber-attacks are expected to change in a changed technological environment if the potential of AI is to be exploited. The exploit of the increased digital attack surface Technologies are becoming increasingly complex, with the result that there is more and more potential vulnerability. In addition, today’s systems are so complex that they will slowly become absolutely not transparent to humans. Just as it does not matter whether a rock gorge needs to be protected with a given number of people, or a stretch of coastline of hundreds of kilometers, the technical complexity can also become such a very long stretch of coastline. Roughly speaking, this can be summed up in the phrase “the digital attack surface is getting bigger”. Not only installing a component, or deleting, or updating become a new vulnerability, but it can also result from using up the system itself. This has been the case for some time, but so far these specific vulnerabilities have been difficult for attackers to find Traditionally, such hidden vulnerabilities have been found either by analysts or by simple testing programs. Two approaches are used, one focusing on known types of errors, the other finding vulnerabilities by analyzing user error phenomena. To date, even attackers have been able to thinking only in the dimensions of test programs and code analysis. But AI opens up a new “third dimension” in vulnerability detection. A teachable algorithm reconnoitres internal errors in complex systems much more efficiently than a human or a traditional program. It is especially effective to use such debugging for specific purposes, as our “semi-skilled worker” can be trained faster and more accurately for a specific system. With this, the reconnaissance ability, ie the preparation for the attack, is very strengthened - and if the weak points are known, the attack will be more efficient and faster. Thus, we can be imagined such an attack in which an aforementioned “semi-skilled worker” AI performs his reconnaissance work, and other AI modules use the information obtained to perform

American Journal of Research, Education and Development 37 ISSN 2471-9986 2019/4 the attack task [8]. If we add to this the fact that AI is also very successful in other areas (camouflage, targeting, etc.)4, adding any number of modules to the method can further increase the effectiveness of the attack. The treatments of vulnerabilities are needs to be reconsidered. Nowdays the vulnerability-management is addressed through various databases (CVE, NVD, US-CERT, Snyk, etc.). These databases store problems and their fixes, to keep the lot of programs and tools safe. But these databases can also be used to train offensive AI systems. In a simpler case, they only attack more optimally with the knowledge thus obtained [9]. But if a more advanced AI can handle the vulnerabilities of a database as samles - it will also recognize vulnerabilities similar to those in the database, but hitherto unknown5. Based on these, it is suggested to rethinking the current system of publicity of such a database. This is because the method that has so far helped increase security through AI, now can be turned against security. The existence of AI can also be exploited. The AI is dangerous, not only when they are using it for performing an attack. What if the enemy acquires the set of data with which we have trained our own AI system? This datas can be used for abuse in many ways: they gives an attacker an idea of us, they can be used on to train its own AI system, but they can also be used for classic cyber methods, e.g. they can get tips from personal data to crack your password. They can even be used outside cyberspace, e.g. an attacker can use sensitive information in the database to blackmail a prominent person. Nor can it be ruled out that the entire trained AI will be stolen and further developed, turned against us. It’s like blowing up a gunpowder-depot at a castle siege: it requires treason, but it’s not impossible - and what had hitherto helped or protected is now destroying the defenders. Attacks will be more sophisticated and personalized The artificial learning of speech and habits, as well as the development of image recognition and imaging, have opened up new perspectives in modern deception, so-called social engeneering (soen - in other words, psychological influence). From our point of view, we can divide this form of 4 We will discuss this in more detail in the second part of our article when analyzing DeepLocker. 5 Of course, this method is also used by defense professionals, but they patch the new vulnerability, if the system finds one - instead of using it for attacking.

American Journal of Research, Education and Development 38 ISSN 2471-9986 2019/4 digital fraud in two. The breakdown below is based on AI-based attacks used nowday, the divisions in the soen- literature are different [ 10]. 1. One part of the soen is based on digital imitation of reality, we called “imitation deception”. Among other things, they use, for example, that technique for deception, which serves the perfect CGI movies and games entertainment. 2. In the other part of the \"soen\", the key is increasingly perfect mapping of the system of people’s personal, unique habits, called “research scams”. This segment is similar to what the literature refers to as the information gathering phase of the \"social engeneering” [3]. For the latter point, let’s take a spam-based scam as example. Everyone has come across that they want to persuade us to open a (viral) attachment or website, or provide our private information. Many people have already come across chat robots, which e.g. used by a customer services to automatically answer typical problems (these also use AI). Well, if you use such a chat-bot-like, teachable program for sending emails, you don’t need to attack the target for the first time - as fewer and fewer fall for these attempts. But the spam robots are able to build a trust with a few mail exchanges – this is an ancient basic trick of and swindlers! The victim then unsuspectingly walks into the trap sent by the “kind customer service representative” (actually an AI). This way, data can also be tricked out of victims, who thinks they are giving them to a real company or person. Or this way they will download a viral file, disguised as a personal offer made for them, and so on. All this automatically million times… To feel how advanced the meaningful text produced by the AI is, check out this marketing blog doesnotexist.com. It’s a marketing blog at first glance, but it’s actually written by AI. According to some, this technology can ruin the good functioning of present search engines by producing pseudo-content [11]. An example of the use of \"imitation deception\" would be, if an attacker artificially imitate the voice and speech of a bank manager, and thus give verbal instructions for a false transfer. Moreover, the same can be done with video calling. If we watch videos about the capabilities of American real- time face-changing software called face2face6 [12], it becomes clear that the person, who we think we see and hear can be anyone! Identity theft is becoming easier, free programs have been available for years for this [13]. While this is not the main technology of cyber attacks, it can still provide an opportunity for the most spectacular scandal in cyberspace if e.g. such fake video is used for political 6 A spectacular video of the technology is also available: https://www.youtube.com/watch?v=ohmajJTcpNk

American Journal of Research, Education and Development 39 ISSN 2471-9986 2019/4 manipulation on social media sites. Or even such a \"recording\" can be used for alibi production - and criminals can have many more ideas... Because personality theft more than cyber threats and is a real-world problem, AI-supported personality theft is also only partially a cyberspace problem, it's a much more complex source of threat. Asymmetric attacks The companies ’cyber defense professionals have adopted the term asymmetric warfare from the military terminology. It is used in the sense that while the defense must be 100% effective against a plethora of attacks, for the attacker it is enough to be successful once, out of countless attempts. That is, a million successful attack defenses are worthless if a single intrusion succeeds. Asymmetry, of course, doesn’t give the attacker a million-fold advantage, a single soldier does not have to fight against a million weapons. Rather, we can imagine that e.g. a single box of ant powder that is sprinkled on the ground, against enough millions of ants. But if an ant comes in through the ceiling, it marks the way, the others come after it - and a whole new way has to be figured out how they can’t go on the ceiling either. AI helps to enumerate such opportunities: e.g. figuring out that attackers can come in through the ceiling, he examines whether dust scatter is certainly not applicable for defense. But in response to AI based defending of the companies, the attackers are using the same technology, so the asymmetry of combat does not change. The rate may drop, perhaps “only” a thousand times the chances of an attacker - but the disproportion remains, that to get in once is enough. AI cannot create effective defense either, as the attacking AI will always find new ways, will be able to bypass the defense. So there will still be a need for human creativity, intuition and attention. If, on the other hand, the defending party stays with the classic defense, the chances of the attacker increase so, that the asymmetry grows almost to infinity. A planned and AI-based attack against a system protected by a classic firewall and antivirus will be just as indefensible as a leather shield against a helicopter machine gun. Completely new, completely AI-based methods will be formed Many workshops are constantly studying the development of technologies to identify and predict new threats, and to be a few steps ahead of cybercriminals. One such fundamental novelty is program operation based on biological functions. This has a long history, as we do not have to pay

American Journal of Research, Education and Development 40 ISSN 2471-9986 2019/4 royalties for the improvements that nature has made over millions of years, we are free to draw ideas from it. With AI coming to the fore, much more can be copied than ever before. Already a separate science, biotics deals with the biological mechanisms that can be utilized in programming models. In cyberspace, perhaps the possibility of swarm viruses is one of the most interesting areas. These models represent a branch of AI research, as they are not controlled by a learning brain, but by the joint work of “many small brains” the most effective solution is created. The goal is not robot control like a dog’s brain, but a \"living being\" embodied by an anthill or a swarm of wasps. For such communities, the boly, swarm, herd (etc.). represents a distinct quality of life, and the lives or sacrifices of the members are dwarfed in comparison, the combined activities of the members create an effective solution. But there are serious possibilities in the AI-based encryption, that viruses can use to hide better themselves. For example, Deeplocker locks everything (that antivirus technology might perceive as a threat) into a multi-layered, AI-based black box, thus avoiding protection. An example of a swarm virus has been published by a Czech team that exploits the self-improvement method inherent in swarm intelligence. Both solutions are novel in terms of the possibilities of hiding by AI, only in different ways. These two examples will be described in detail in the next part of our series, where we would like to illustrate the novelty of AI-based cyber-attacks with other examples. Previously proven defenses are mathematically insufficient because of AI The adequacy of the protection of a system is always determined by how long it takes to crack it with the available technologies. The constant increase in the computing speed of the machines makes the previous protections always insufficient. Outdated encryption techniques are constantly being discarded7 or tightened. An example of this is a classic method of protection against robotic identifications, the “Captcha caracters,” which will need to be replaced by completely new captcha methods. In the next part of our article series, we present the method of hacking the classic captcha protection by connecting several AI modules, which makes this innovation necessary. In the case of Captcha, therefore, it is sufficient to turn to a different kind of question, this defends us against robots. However, when appears an order of magnitude more efficient technology (thousandfold or millionfold efficient), it does not protect us e.g. the expected password is a few 7 In the next part of our article-series, we present the obsolescence of a classic method of protection against robot identifications, “Captcha fonts”, with a description of their hacking.

American Journal of Research, Education and Development 41 ISSN 2471-9986 2019/4 characters longer and a little more complicated. The advent of quantum computers and the proliferation of AI represent such a leap of magnitude. The combined use of the two could become a cyber weapon that could force any welfare state to the floor in the cyber battlefield, who does not have this technology for defense purposes either. The two extreme attitudes Based on the above, it can be perceived that there is a real threat. But – as is customary be among people – there are those who see this extremaly: many exaggerate, and many downplay the problem. These extreme opinions can also clarify the description of the situation, so it is worth thinking briefly about their arguments. The alarming view A few years ago, the exaggerators were in majority. In 2017, almost all tech-portals took the view, that behind the 62% of cyber attacks would be the AI in 2018 [14]. The source of the fake news was a survey conducted among professionals attending the Black Hat conference, who saw the future so darkly - perhaps as a result of the conference. But their opinions seem to be based on emotions rather than their predictive analytical algorithms. The news and statistical results did not come even in 2019 about the fulfillment of this ominous prophecy, and there was no official data on the side of anti-virus companies documenting a high-risk attack using artificial intelligence. More specifically, we can find news, even on (seemingly) professional sites, that document an AI attack in their title, but in fact these were usual assaults. For example, some authors are already talking about the use of AI in connection with the infamous Trojan virus called Emotet. In the description say artificial intelligence does that the virus's ability to mimic the victim's language more authentically and personalized than the traditional \"click to the attachment\" text. As a result, the victim is more likely to click on the infected email-attachment [ 15]. But there is no other source to confirm that the virus uses AI - in fact, all antivirus-vendors mention it as a dangerous but traditional virus, they do not connote AI, neither Kaspersky nor Symentec report [16]. It also described in the literature as a truly significant and skillful attack, especially the joint activity of the Emotet-Qakbot virus pair. But nobody talks about AI, nor self- developing email language. In connection with another pair of Emotet (called TrickBot), already

American Journal of Research, Education and Development 42 ISSN 2471-9986 2019/4 mentioned the AI, among others Max Heinemeyer8 – but rather in the sense, that impossible defending against it by human intervention, because it spreads so rapidly [17]. There is a consensus that a similar malware can only be curbed by AI-based defense, as Microsoft officially reports, but there is no word on the AI background to the attack [18]. The case just analyzed urges us to treat news about such events with reservations - world hysteria could have erupted if a large news portal treates the case as an AI attack, and overreact it for the sake of viewership. That is what we must expect. The media won’t interest in the percentage of AI in a cyber attack, provoke frighten has been a good deal so far, and politicians can write on thier flag the phrase of AI’s anti-threat as well. On the other hand, undocumentedness does not rule out that those who take the danger too seriously are right, but in this way the topic can only be discussed at the level of conspiracy theories, not scientifically. The reason for the non-existence of such news may be quite simply that no one wants to leak information about the way they were attacked, or were unable to detect such an attack, because were so fast and effective, etc. Skepticism It has to be seen, that if we put scientific arguments on the balance, we can find far fewer arguments on the side of skeptics than we do on frightening ones. Nevertheless, few one are afraid of it - more precisely, the majority is unaware of the danger. If we ask people what global threats they fear, it would come to their mind terror explosions, biological virus infections, or environmental disasters, economic crises. But probably just a few would list this technology, like sci-fi fans and IT professionals9 Skeptics may even bring up that AI is an expensive thing, so it will spread slowly. According to this, AI-based attacks are also costly, so hackers will stay with the cheaper solution and will not even be able to take advantage of the opportuninies opened up by AI. However, this can be thought of only by such journalist, who watch hackers only in TV thrillers, who are sitting in front of their laptop in hoodie, or fighting for the truth in small groups in the cellars of abandoned factories ... But 8 The very interesting study does not mentions AI-based attack, but by analyzing some of attacks, he rethinks what the attacking techniques will be able to do by expanding with AI. 9 I conducted a non-representative, only informative survey among my acquaintances. If we don’t list AI as an optional problem, very few will mention it on their own. If we make it part of a list it, noone put it to the first place either.

American Journal of Research, Education and Development 43 ISSN 2471-9986 2019/4 looking at the facts, unfortunately, they have a lot of money, even for such expensive technology. On the one hand, significant state subsidies (perhaps secret) cannot be ruled out, as this is also a kind of arms race. On the other hand, various groups have huge revenues from cybercrime: according to a study based on 2017 data, the amount received for them was already estimatedat $ 1.5 trillion (thousands-billion, 1012) [19]. Similar data can be found for the coming years, but it is estimated that by 2021, this could reach 6 trillion [20]. Plus, software development isn’t used to start from scratch in such circles either: it’s easier to steal an AI developed for another purpose and teach them what they want. Some AI enhancements are no secret anyway, and there are number of free options available [21]. Primarily for the python programming language, there is already an extensive module system (library) that supports AI functions, or e.g. javascript based TensorFlow developped by google is also available10. The middle-road To sum up the opposition between panic-monger and skeptics, it can be said, that even if we don’t panic about the possibility of AI-based cyber-attacks, we need to be seriously prepared for it, with significant financial outlay. Clearly, the fear based on literary works and films will not be able to provide a solution, just as our proven methods and systems are neither enough. In the past, bronze or leather armor was not suitable against iron weapons, instead iron shields or iron armor provided a chance for effective defense, which in turn was significantly more expensive. Similarly, the “take a hair of the dog that bit you” principle needs to be applied today. Based on this, AI-based defense will be able to compete against AI-based cyber attacks. But just as the Iron Shield did not protect everyone against a cleverly-used iron weapon in all cases, AI-based defense will not be able to provide 100% protection. In addition to acquiring the right technology and training well- paid team of professionals of AI-based protection, it will also be necessary to rethink and enforce security rules more thoroughly. Summary This brief overview was able to outline just a few ways to use AI technology in cyber attacks. Perhaps that was enough to highlight that the danger was great. After a more in-depth analysis of 10 https://www.tensorflow.org/js/

American Journal of Research, Education and Development 44 ISSN 2471-9986 2019/4 defense systems, we can be hopeful that this challenge is not impractical if we also use AI in defense (this is the subject of another study in our study). Perhaps there are also greater or more spectacular dangers of AI than those caused by cyberspace, such as autonomous combat systems. But this is not insignificant either, because if we are not able to control the cyber world with the right countermeasures, this could be ending the open internet, telecommunications or television. Namely this source of danger can ultimately be reduced to zero, if we are switching off the convenience electronics completely or partially. The question, then, is whether world society will be able to prevent this from happening - and other hyper-mod-ern technologies (genetic, nano, atomic or otherwise) will not become the enemy of humanity instead of aid.

American Journal of Research, Education and Development 45 ISSN 2471-9986 2019/4 REFERENCES [1.] Négyesi I. A mesterséges intelligencia és a hadsereg I. Budapest: Hadtudományi Szemle X/2. 2017. 24-28. [2.] Russel S, Norvig P. Mesterséges Intelligencia Modern megközelítésben. Budapest: Panem; 2005. 1-4. [3.] Kovács L. A kibertér védelme. Budapest: Dialóg Campus; 2018. [4.] Csirmaz E. LAWS – a gyilkos robotok és a nemzetközi jog. November 8, 2018. https://arsboni.hu/laws-a-gyilkos- robotok-es-a-nemzetkozi-jog/. Accessed June 13, 2019. [5.] Bodnár Z. A magyarok megállítanák a gyilkos robotokat, a nagyhatalmak ragaszkodnak hozzájuk. June 07, 2019. https://qubit.hu/2019/06/07/a-magyarok-megallitanak-a-gyilkos-robotokat-a-nagyhatalmak-ragaszkodnak- hozzajuk. Accessed July 13, 2019. [6.] EP 2018/2752(RSP) Resolution on autonomous weapon systems Europa.EU. September 12, 2018. https://www.europarl.europa.eu/doceo/document/TA-8-2018-0341_HU.html. Accessed June 12, 2019. [7.] Eidson B. Creating an AI Red Team to protect critical infrastructure. September 2019. https://www.mitre.org/publications/project-stories/creating-an-ai-red-team-to-protect-critical-infrastructure. Accessed July 19, 2019. [8.] Chartis-Research. The State of AI in Risk Management. October 22, 2019. https://www.chartis- research.com/technology/artificial-intelligence-ai/state-ai-risk-management-10976. Accessed July 21, 2019. [9.] Fu J. Security: Using AI for Evil. June 11, 2018. https://threatvector.cylance.com/en_us/home/security-using-ai- for-evil.html. Accessed June 27, 2019. [10.] Haig Zs. Információs műveletek a kibertérben. Budapest, Dialóg Campus Kiadó, 260-261. Budapest: Dialóg Campus; 2018. [11.] Vincent J. Endless AI-generated spam risks clogging up Google’s search results A ‘tsunami’ of cheap AI content could cause problems for search engines. July 02, 2019. https://www.theverge.com/2019/7/2/19063562/ai-text- generation-spam-marketing-seo-fractl-grover-google. Accessed July 27, 2019. [12.] Thies J, Zollhöfer M, Stamminger M, Theobalt C, Nießner M. Face2Face: Real-time Face Capture and Reenactment of RGB Videos. July 2016. http://www.graphics.stanford.edu/~niessner/papers/2016/1facetoface/thies2016face.pdf. Accessed May 29, 2019. [13.] Tran D. Face2face — A Pix2Pix demo that mimics the facial expression of the German chancellor Jul 4, 2017. July 04, 2017. https://towardsdatascience.com/face2face-a-pix2pix-demo-that-mimics-the-facial-expression-of- the-german-chancellor-b6771d65bf66. Accessed June 19, 2019. [14.] Black Hat Research. AI as an Attack Method. August 01, 2017. https://isssource.com/black-hat-ai-as-an-attack- method/?fbclid=IwAR3No6HCNJ8BJsjUGgnXwxKBUXLe1trqSG1wGbBSynFBiIJWunsU7-XuV30. Accessed May 25, 2019.

American Journal of Research, Education and Development 46 ISSN 2471-9986 2019/4 [15.] Dixon W, Eagan N. 3 ways AI will change the nature of cyber attacks. World Economic Forum. June 19, 2019. https://www.weforum.org/agenda/2019/06/ai-is-powering-a-new-generation-of-cyberattack-its-also-our-best- defence/. Accessed June 03, 2019. [16.] Symantec. The Evolution of Emotet: From Banking Trojan to Threat Distributor. July 18, 2018. https://symantec-enterprise-blogs.security.com/blogs/threat-intelligence/evolution-emotet-trojan-distributor. Accessed June 05, 2019. [17.] Heinemeyer M. The Next Paradigm Shift - AI-Driven Cyber-Attacks. DarkTrace Research White Paper, 2. 2018. https://pdfs.semanticscholar.org/6b18/6268d00e891f3ed282544ac5833c01a2891c.pdf. Accessed June 03, 2019. [18.] Microsoft Defender ATP Research Team. How artificial intelligence stopped an Emotet outbreak. February 14, 2018. https://www.microsoft.com/security/blog/2018/02/14/how-artificial-intelligence-stopped-an-emotet- outbreak/). Accessed June 08, 2019. [19.] Dr. McGuire M. Into The Web of Profit An in-depth study of cybercrime, criminals and money. Cupertino: Bromium Inc.; 2018. [20.] Morgan S. Official Annual Cybercrime Report. Toronto: Herjavec Group; 2019. [21.] Nanalyze. A List of 15 Free AI Software Programs to Download. March 16, 2017. https://www.nanalyze.com/2017/03/free-artificial-intelligence-ai-software. Accessed June 20, 2019.