Deep Learning Framework for Cyber Attack Prediction
Added on 2023-02-13
12 Pages2842 Words59 Views
Deep Learning Framework for
Cyber Attack Prediction
Literature Review (E1)
Research Methods
Date:
Supervisor:
Name:
Student ID:
Cyber Attack Prediction
Literature Review (E1)
Research Methods
Date:
Supervisor:
Name:
Student ID:
Table of Contents
1. Introduction .............................................................................................. 1
2. Literature Analysis ..................................................................... 2
3. Conclusions .............................................................................. 7
References .................................................................................. 9
1. Introduction .............................................................................................. 1
2. Literature Analysis ..................................................................... 2
3. Conclusions .............................................................................. 7
References .................................................................................. 9
1
1. Introduction
“Cyberattack” is a hot topic in today’s digitalized world. Cyber technologies have opened
new doors to organizations and people to digitize them with the latest technologies and
ease their work. Unfortunately, this has opened opportunities for illegitimate users like
hackers as well. Cyber attacks are increasing immensely and make a great loss to
organizations. For example, cyber-attacks increased by 37% over the past month with
COVID19 pandemic (Muncaster, 2020). According to Bloomberg, a cyber attack hit the
U.S. Health Agency this month to steal national health data and to publish fake
news(SteinJacobs, 2020). In January 2020, the Puerto Rico government lost $2.6
Million in a phishing scam(PRESS, 2020). These cyber-threats make monetary losses,
loss of trust, harm to people and it directly affects the reputation of the companies.
There are a lot of cybersecurity tools out there to secure cyberspace. It is always a
necessity to predict the attack in the early stages before it harms the organization. So
that the companies can set up precautions to stop the attack. For example, Honeypots
and network telescopes monitor unsolicited internet traffic to gather data related to
cyberattacks in networks and to allocate defence tools to secure the network(Peng et
al., 2016).
The subject of predicting cyber attacks and efficient models have reached a top
research topic in the last few years. Different authors followed different approaches to
predict cyberattacks while having advantages and disadvantages in their models in
different situations. The current research path moving towards top-notch technologies
like artificial intelligence, machine learning, deep learning and big data.
1. Introduction
“Cyberattack” is a hot topic in today’s digitalized world. Cyber technologies have opened
new doors to organizations and people to digitize them with the latest technologies and
ease their work. Unfortunately, this has opened opportunities for illegitimate users like
hackers as well. Cyber attacks are increasing immensely and make a great loss to
organizations. For example, cyber-attacks increased by 37% over the past month with
COVID19 pandemic (Muncaster, 2020). According to Bloomberg, a cyber attack hit the
U.S. Health Agency this month to steal national health data and to publish fake
news(SteinJacobs, 2020). In January 2020, the Puerto Rico government lost $2.6
Million in a phishing scam(PRESS, 2020). These cyber-threats make monetary losses,
loss of trust, harm to people and it directly affects the reputation of the companies.
There are a lot of cybersecurity tools out there to secure cyberspace. It is always a
necessity to predict the attack in the early stages before it harms the organization. So
that the companies can set up precautions to stop the attack. For example, Honeypots
and network telescopes monitor unsolicited internet traffic to gather data related to
cyberattacks in networks and to allocate defence tools to secure the network(Peng et
al., 2016).
The subject of predicting cyber attacks and efficient models have reached a top
research topic in the last few years. Different authors followed different approaches to
predict cyberattacks while having advantages and disadvantages in their models in
different situations. The current research path moving towards top-notch technologies
like artificial intelligence, machine learning, deep learning and big data.
2. Literature Analysis
This section introduces the related work conducted in the past by different researchers.
This section includes different approaches they used, their conclusions, findings and
drawbacks.
The paper written by Ling, introduces a regression-based analytical model to predict
cyber attacks in Honeynets. Honeynets are composed of multiple honeypots. The main
purpose of conducting this study was to identify geospatial and temporal patterns in the
cyber attacks and use the knowledge for future attack predictions. The authors
introduced a vector autoregression(VAR) model for honeynets. This study used a
dataset with 9 AWS virtual honeypot hosts. The paper well explained the ways to use
VAR and BigVAR models in predictions. The researchers introduced a fractional
integration methodology to calculate Large Range Memory- LRM in each host which
helps to achieve concise modelling and high performance. Moreover, they found that
the dependency among hosts does not improve prediction accuracy if honeypot hosts
are not directly connected in the same network(Ling et al., 2019).
The research conducted by Peng(2019) about cyberattack rates used another approach
to accurately predict cyber attacks. These authors have studied extreme value
phenomenon exhibits in cyberattack rates. In short, extreme value phenomenon is the
number of attacks against a system of interest per time unit. This value is very important
when allocating defence resources by the defender to protect networks at the right time.
The authors introduced a marked point process technique to model and predict these
extreme cyber-attack rates. They have used Value-at-Risk(VaR) to measure the
intensity of the attacks over a period of time by providing the probability of extreme
cyber-attack rates with a certain confidence level. In addition to that, they have used
the Point Over Threshold(POT) method to model the magnitude of extreme attack rates.
This section introduces the related work conducted in the past by different researchers.
This section includes different approaches they used, their conclusions, findings and
drawbacks.
The paper written by Ling, introduces a regression-based analytical model to predict
cyber attacks in Honeynets. Honeynets are composed of multiple honeypots. The main
purpose of conducting this study was to identify geospatial and temporal patterns in the
cyber attacks and use the knowledge for future attack predictions. The authors
introduced a vector autoregression(VAR) model for honeynets. This study used a
dataset with 9 AWS virtual honeypot hosts. The paper well explained the ways to use
VAR and BigVAR models in predictions. The researchers introduced a fractional
integration methodology to calculate Large Range Memory- LRM in each host which
helps to achieve concise modelling and high performance. Moreover, they found that
the dependency among hosts does not improve prediction accuracy if honeypot hosts
are not directly connected in the same network(Ling et al., 2019).
The research conducted by Peng(2019) about cyberattack rates used another approach
to accurately predict cyber attacks. These authors have studied extreme value
phenomenon exhibits in cyberattack rates. In short, extreme value phenomenon is the
number of attacks against a system of interest per time unit. This value is very important
when allocating defence resources by the defender to protect networks at the right time.
The authors introduced a marked point process technique to model and predict these
extreme cyber-attack rates. They have used Value-at-Risk(VaR) to measure the
intensity of the attacks over a period of time by providing the probability of extreme
cyber-attack rates with a certain confidence level. In addition to that, they have used
the Point Over Threshold(POT) method to model the magnitude of extreme attack rates.
End of preview
Want to access all the pages? Upload your documents or become a member.