Evaluating the Provision of Botnet Defences using Translational
Research Concepts
Dilara Acarali, Muttukrishnan Rajarajan and Nikos Komninos
School of Mathematics, Computer Science and Engineering, City, University of London, London, U.K.
Keywords:
Cyber-security, Botnets, Translational Research, Implementation Science, Botnet Propagation.
Abstract:
Botnet research frequently draws on concepts from other fields. An example is the use of epidemiological
models when studying botnet propagation, which facilitate an understanding of bot spread dynamics and
the exploration of behavioural theory. Whilst the literature is rich with these models, it is lacking in work
aimed at connecting the insights of theoretical research with day-to-day practice. To address this, we look
at botnets through the lens of implementation science, a discipline from the field of translational research in
health care, which is designed to evaluate the implementation process. In this paper, we explore key concepts
of implementation science, and propose a framework-based approach to improve the provision of security
measures to network entities. We demonstrate the approach using existing propagation models, and discuss
the role of implementation science in malware defence.
1 INTRODUCTION
Botnets are malware-based platforms built illicitly on
vulnerable networks to serve a cyber-crime agenda. In
botnet research, propagation modelling often appro-
priates epidemiological models of diseases (Brauer,
2008) to identify dissemination factors, to devise im-
munisation strategies (Yong et al., 2012), and to pre-
dict reach and speed. However, findings may not
be applied appropriately in day-to-day practice. Net-
works are highly variable, meaning that any proposed
security measure will fit some scenarios but fall short
in others. Translational research (TR) is a field of
public health care for converting experimental results
into care for patients (Rubio et al., 2010) (Woolf,
2008). We believe that botnet research would benefit
from a similar approach. Standard methods for ob-
serving and evaluating security provisions would en-
sure that a) they are fully utilised in practice, and b)
they are applied consistently and effectively for max-
imum impact.
In this paper, we aim to start a discussion about us-
ing TR and implementation science (IS) to effectively
apply botnet research findings to real-life networks.
This is particularly relevant to botnet propagation,
where mitigation can prevent worsening outcomes.
To our knowledge, this topic has not previously been
explored. Our contributions are 1). an adaptation
of the Dynamic Sustainability Framework (DSF) to
apply IS methods to botnet defence, 2). an analysis
of propagation-based usage scenarios, demonstrating
how DSF can help deliver better protection, and 3). a
discussion of the potential role of IS in malware de-
fence. Section 2 provides a background on TR. Sec-
tion 3 introduces the DSF and demonstrates its use,
whilst Section 4 provides a discussion on the usage of
IS. Related works are covered in Section 5, and we
conclude in Section 6.
2 BACKGROUND
Translational research (TR) investigates the process
of converting scientific knowledge into practical solu-
tions for standard practice, described as “the interface
between basic science and clinical medicine” (Woolf,
2008). Basic science is the pursuit of knowledge with
no practical consideration (Rubio et al., 2010) (analo-
gous to bot case studies). Clinical research introduces
patients for behavioural analysis (Rubio et al., 2010)
(analogous to botnet propagation simulations). TR
has 2 parts (Figure 1). The first (called T1) applies
lab-based studies to clinical trials, whilst the second
(T2) uses obtained results to aid practical decision-
making (Rubio et al., 2010) (Woolf, 2008). T1 re-
quires in-depth scientific knowledge and the relevant
tools for innovative research, whilst T2 requires un-
derstanding of communities, culture, and work envi-
514
Acarali, D., Rajarajan, M. and Komninos, N.
Evaluating the Provision of Botnet Defences using Translational Research Concepts.
DOI: 10.5220/0006884405140519
In Proceedings of the 15th International Joint Conference on e-Business and Telecommunications (ICETE 2018) - Volume 2: SECRYPT, pages 514-519
ISBN: 978-989-758-319-3
Copyright © 2018 by SCITEPRESS Science and Technology Publications, Lda. All rights reserved
Figure 1: Breakdown of translational research into sub-
fields.
ronments (Woolf, 2008) (Rubio et al., 2010). T2 has
2 stages: 1) the translation of results to practice, and
2) the evaluation of this process (Woolf, 2008), also
known as implementation science (IS). For botnets,
we focus on T2, with the development, application,
and evaluation of security policies.
To reap maximum benefit from T1 results, there
must be a systematic, well-observed process of im-
plementing new ideas into existing systems (Woolf,
2008) (Khalil, 2016) (Bauer et al., 2015). IS broadly
refers to the application of new approaches (Woolf,
2008) with performance evaluation and due consider-
ation of surrounding influential factors (Khalil, 2016).
A single new method is an intervention, and a pack-
age of interventions (with a common goal) is a strat-
egy (Bauer et al., 2015). IS considers the possibility
of inconsistent application leading to “quality gaps”
(Bauer et al., 2015), and as a result, enables us to ob-
serve and evaluate natural variation and responses to
interventions (Bauer et al., 2015). Health care litera-
ture organises this process into frameworks.
3 FRAMEWORK
3.1 Outline & Definitions
We use the Dynamic Sustainability Framework
(DSF), proposed by Chambers et al. (2013). It incor-
porates internal and external context environments, as
well as temporal change, allowing us to evaluate the
intervention over time. This means that DSF focuses
on sustainability to address “voltage drops” and “pro-
gram drift” (Chambers et al., 2013). It enables the
continuous improvement of results by exposing the
intervention to changing populations, and adapting
it to better fit emerging conditions (Chambers et al.,
2013). In networks, segmentation results in internal
and external spaces. User behaviours, business prac-
tices, and new technologies add variability that can
affect intervention roll-out. These factors impact the
uniform implementation of new policies, highlighting
the need for spatio-temporal sustainability. Evolving
malware also means interventions must be robust and
adaptable.
DSF splits the process into 3 layers; the interven-
tion, the practice setting, and the ecological system
(Chambers et al., 2013). We split the practice setting
into 2 parts for a layered inner environment to mimic
the layers of an organisational network (Figure 3). In-
fluencing factors at each layer are defined as a set of
constructs. Our adaptations of these are listed in Table
1. The intervention includes actionable steps, desired
outcomes, delivery platforms, and practitioners. It is
deployed within the practice setting, which represents
all internal influences (e.g. systems, resources, and
staff) (Chambers et al., 2013). This sits within the
ecological system, representing external factors like
legislation, regulation, and market forces (Chambers
et al., 2013). DSF operates over periods, denoted as
(T
0
, T
1
, ... T
n
). Hence, the intervention process cov-
ers each layer across all periods, revealing changes in
implementation. This approach is unique to DSF, de-
signed to improve patient health by adapting interven-
tions as required (Chambers et al., 2013). Observing
how the network state changes with time under our
defensive interventions is key to botnet mitigation.
3.2 Usage Scenarios
We now outline how the framework may be applied
to impede botnet propagation. Three epidemic sce-
narios are considered, in: 1) an enterprise network,
2) a mobile network, and 3) a social network. We
use epidemics as our practical example because prop-
agational success plays a large role in determining
a botnet’s eventual impact. The propagation stage
also provides opportunities for mitigation of future
attacks. Additionally, epidemic modelling is widely
used in botnet literature to test behavioural theories
and spread parameters. Lastly, epidemic modelling
fits into the clinical research stage of T1, and we can
carry the results over to the real-life settings of T2.
3.2.1 Enterprise Network Epidemics
Yong et al. (2012) used a modified SIRS model to
define reproductive ratio R
0
in relation to 2 types of
I (infected) node; hidden and active. They recom-
mended keeping R
0
< 1 by a) minimising S (suscep-
tibles), b) increasing removal rate, and c) decreas-
ing transmission rate. Success depends on defence
architecture, location, and deployment time. Dagon
et al. (2006) used the SIR model to capture the di-
urnal nature and regional bias of botnets. Observ-
ing a “natural quarantine effect” (Dagon et al., 2006),
Evaluating the Provision of Botnet Defences using Translational Research Concepts
515
Figure 2: Diagram of updated Dynamic Sustainability Framework (Chambers et al., 2013), showing the 4 layers and their
constructs, along with increments in time (T
x
).
they found that worms have optimal release times
and suggested appropriate surveillance and geograph-
ical/temporal prioritisation for defenders.
We now apply the framework to bolster defences
against a possible epidemic outbreak. Based on the
results above, this network’s intervention includes im-
proving user awareness and patching policies (to min-
imise S), reducing R
0
, identifying risky malware de-
ployment spots and times, centralising security sys-
tems across departments, and deploying a resilient
infrastructure (for quarantining to minimise contact).
The patients are network nodes, and practitioners are
the users, admins, policy makers, and security direc-
tors. The 1st internal context is a network segment
(e.g. a departmental LAN), whilst the 2nd is the wider
enterprise, which is made up of multiple segment con-
texts. Meanwhile, the external context is the con-
nected community around this enterprise. For evalu-
ation, we need to consider the quality of training and
user receptiveness, adherence to policies, and system
performance. Collecting data on these factors then
provides a feedback loop, allowing both the strategy
and the implementation to be improved (Chambers
et al., 2013).
3.2.2 Mobile Network Epidemics
The SIR model is used by Khouzani et al. (2012) in
their proposed malware quarantine scheme based on
regulating each node’s communication range. When
the reception gain of S nodes is decreased, contact fre-
quency also drops. This should be balanced against
the lowest acceptable QoS for the network. Lu et al.
(2016) employs epidemic principles stochastically to
demonstrate that spread in WiFi networks is depen-
dent on node proximity, and specifically on node den-
sity, wireless transmission range, and the node mobil-
ity radius. Propagation eventually stops if the mobil-
ity radius is limited enough (Lu et al., 2016).
This time, an epidemic is already underway. The
intervention involves responsive measures like track-
ing of I nodes, grounding (where possible) of mobile
devices (reducing their mobility radius), reducing the
gains of S nodes near to I nodes, and performing tar-
geted clean-up. Users should be notified and may as-
sist in grounding. As before, nodes represent patients,
whilst users, admins, and owners of WiFi infrastruc-
ture are the practitioners at various levels. The 1st
internal context is the specific cell or geographic lo-
cation of the infection, whilst the 2nd is the wider cel-
lular/geographic region surrounding it. The external
context is the wider WiFi network, and the Internet.
The effectiveness of this implementation will be im-
pacted by users’ attitudes to the event and to the mea-
sures taken. For example, the reduction in network
quality might garner a negative reaction. The prac-
ticality of the grounding policy, ability to accurately
detect and track I nodes, and the efficiency of the re-
covery process also need to be measured for evalua-
tion, and can provide opportunities for improvement.
3.2.3 Social Network Epidemics
Sanzgiri et al. (2012) used the SI setup to model
Twitter-based botnets, with propagation impacted by
degree of followers, activity levels, and the level of
response/interest in tweets, contributing to the proba-
bility of user clicks on malicious URLs. They sug-
gest limiting the sharing of links, or better control
and monitoring of short-URLs, noting that individ-
uals may not be aware of potential risks. Yan et al.
(2011) used epidemiological concepts to demonstrate
that propagation rate is influenced by friendship net-
works, activity, and the infection source. They sug-
SECRYPT 2018 - International Conference on Security and Cryptography
516
Table 1: Outline of network constructs at each layer of updated DSF (Chambers et al., 2013).
Constructs Description
Intervention
Components Collection of individual elements that make up the intervention.
Practitioners Individuals who will implement the intervention across all levels.
Characteristics Collection of descriptive characteristics for the intervention.
Aims Targeted outcomes of the intervention.
Delivery Vectors Means of delivery of the intervention.
Practice Setting
Context 1 - Local Network
Infrastructure Topology and design of the network at this level, including defences.
Systems Collection of specialised systems or technical resources used.
People Collection of local staff, including count, roles, and hierarchy.
Culture & Climate Social environment created by management, individuals, and group attitudes.
Business Function Function of this context in the wider network.
Training Nature and availability of knowledge and skills transfer to practitioners.
Supervision Collection of regulators and enforcement methods.
Context 2 - Wider Network
Infrastructure Topology and design of the wider network, interconnection of local contexts.
Systems Collection of systems or resources used organisation-wide, e.g. email services.
People Collection of management staff, including count, roles, and hierarchy.
Culture & Climate Social environment created by management based on brand/organisation ethos.
Business Model Goals of organisation, plus functions/services, financial and implementation plans.
Digital Ecosystem
Regulation & Legislation Industry standards and national legislature influencing operations.
Market Trends Current popular technologies, activities of competitors.
Populations Characteristics of populations engaging with devices and services.
Usage Culture Attitudes towards technology and current trends in behaviour.
Partners Collaborating third-parties, shared infrastructure.
Upcoming Technology Collection of technologies in the pipeline for mass deployment.
gest the use of an early warning system that prop-
agates alerts amongst users to reduce susceptibility,
and the use of centralised servers to monitor and sani-
tise suspicious URLs (Yan et al., 2011).
To defend a social network against a potential epi-
demic, the intervention incorporates early detection,
suspension of infected accounts to reduce activity, in-
creased monitoring for URLs shared by highly-active
highly-connected users, and an in-built URL short-
ening service. Users should be notified of risks and
ongoing security events. The internal context is the
user community or friendship group, encapsulated by
the wider social network, made up of multiple simi-
lar communities. The external context is the rest of
the Internet. Patients could be user accounts or users
themselves, with practitioners being the platform ad-
mins and security staff. Evaluation needs to consider
the culture/brand of the social networking platform,
user engagement with security advice, public recep-
tion to new policies (including privacy concerns), and
the effectiveness of early detection systems. Sug-
gested data collection includes user feedback, activity
logs over time and region, frequency and distribution
of URL and short-URL use, and performance metrics
for detection systems and URL-shortening services.
4 DISCUSSION
The scenarios in Section 3.2 show how the framework
can be applied to different situations, populations, and
infrastructures. It enables us to identify specific ac-
tions via the intervention strategy, and then to deter-
mine areas where implementation may be insufficient.
The provided suggestions are based on literature and
experience, but real-life application will flag signif-
icant, unpredicted problem areas. Consideration of
these factors over time means that improvements can
be made to the intervention incrementally, providing
sustained long-term solutions. Being able to achieve
this with a standardised framework is highly benefi-
cial, as it eases the process and makes different in-
terventions comparable. Sustainability of an inter-
vention should be a top priority. Networks become
vulnerable due to changes in technology and usage
over time, a process mirrored by evolving malware
threats. Therefore, it is difficult to predict what the
future will bring. Furthermore, when deploying new
technology, it is not possible to consider a priori ev-
ery scenario surrounding its use. Hence, interventions
must be flexible enough to incorporate new insights.
The aim of DSF is the improvement of interventions
Evaluating the Provision of Botnet Defences using Translational Research Concepts
517
Table 2: DSF applied to 3 usage scenarios, with interventions (Strategy) derived from epidemic modelling results. Impact
Factors determine intervention success, and Measures denote evaluative considerations.
Scenario 1 Scenario 2 Scenario 3
Settings -Enterprise -Mobile -Social
Strategy -Train users for response -Inform users of situation -Add early detection systems
-Shrink S via patching & updates -Ground infected devices -Send regular advice to users
-Centralise defensive systems -Drop reception gains of close Ss -Suspend infected accounts
-Find optimal spread spots, times -Targeted cleaning of detected Is -Track URLs of active users
-Add network contingencies to
take segments offline
-Add in-built short-URL service
Impact -Engagement of users in training -Culture & attitudes of users -Culture & attitudes of users
Factors -Uniform delivery of training -Ability to ground devices -Reception of new policies
-Adequate supervision -Ability to monitor movement -User response to security advice
-Adherence to patching policy -Efficiency of gains reduction -Effectiveness of early detection
-Efficiency of recovery process -Efficiency of recovery process -Uptake of short-URL service
-Robustness of contingencies
Measures -User skills levels, experience -User experience -User feedback
& -Vulnerability status -Current QoS -Activities over time & region
Feedback -Usage of spread spots -Current estimated S count -Freq. & distribution of URL use
-Activity at optimal times -Current estimated I count -Detection system performance
-Defence system performance -Current R count -Short-URL service performance
- through constant measurement from a dynamic con-
text, interventions can be optimised (Chambers et al.,
2013).
In both health care and cyber-security, a lot of
resources are put into T1. However, T2 efforts
may actually have the larger practical impact (Woolf,
2008). Therefore, the proper translation and delivery
of new research should receive greater funding and
focus. Well-realised T2 endeavours can bring return-
on-investment for T1 processes (Woolf, 2008). Ad-
ditionally, a more pragmatic outlook may be benefi-
cial in epidemiological studies. Galea (2013) argues
that epidemiology should shift to become more con-
sequentionalist (where models’ worth are measured
based on results rather than theory), suggesting that,
given limited time and resources, this would help to
prioritise actions to maximise overall health (Galea,
2013). Finally, TR is concerned with the overall im-
provement of public health, including the observed
environment, plus the overall distribution of health
across different regions (Galea, 2013). We suggest
that a similar mentality be adopted against botnets.
The threat is worldwide and successful defence in one
environment is insufficient - botnets formulated else-
where can still attack this environment. Therefore,
we should work collaboratively on global systems
to mitigate bot malware at the highest level, collec-
tively protecting national systems, businesses, homes,
NGOs, and the shared Internet infrastructure. Only
when this distribution of sustained network health can
be achieved will the botnet threat truly lose its po-
tency.
5 RELATED WORK
Dedeke (2017) discussed the NIST Cyber-Security
Framework (CSF) (NIST, 2014), designed for organ-
isations to identify and reduce security risk within
their systems. CSF aims to create a paradigm shift
from compliance-based to risk management-based
defence, which should yield higher quality precau-
tions (Dedeke, 2017). CSF has 3 sections, with the
core housing 5 functions. These are 1). identification
of requirements, 2). development of safety measures,
3). monitoring, 4). development of action plans, and
5). deployment of recovery strategies (Dedeke, 2017)
(NIST, 2014). This aligns with some actions in our
suggested intervention strategies. CSF outlines a co-
ordination scheme between executive, business and
implementation levels (NIST, 2014), but does not in-
corporate dynamic delivery contexts. Dedeke (2017)
highlights the use of implementation tiers to track
progress over time, as changes are not otherwise con-
sidered.
Bassam and Deborah (2010) presents an ESM
(Enterprise Security Management) framework for
building secure, well-structured enterprises with
adaptive processes. The framework consists of 12
steps, starting with an analysis of requirements, fol-
lowed by: identification of capability gaps, priori-
tisation of tasks, development of architecture, mon-
itoring, and continuous realignment - described as
the “periodic reassessment of requirements, capabili-
ties and updating the architecture” (Bassam and Deb-
orah, 2010). This allows the system to keep pace
SECRYPT 2018 - International Conference on Security and Cryptography
518
with emerging threats. Layers within the enterprise
are considered, with emphasis on information sharing
across domains (Bassam and Deborah, 2010). The
inclusion of change, and the layered view of enter-
prises, makes ESM similar to DSF. However, DSF is
not enterprise-specific, allowing it to be used flexibly
in many different scenarios.
These frameworks approach security implemen-
tation predominantly from a business perspective.
However, they do not formalise a process for turn-
ing pure research (T1) into applicable methods (T2).
As such, they do not focus on getting the best out
of available knowledge. They also do not integrate
time and internal/external contexts, and so cannot im-
prove interventions against changing trends and pop-
ulations. We believe that taking inspiration from the
well-established field of health care provides a unique
angle to exploit in generating new ideas for the main-
tenance of network health.
6 CONCLUSIONS
In health care, TR delivers research knowledge to
patients. We have proposed that a similar approach
be applied for botnet mitigation - bringing technical
knowledge and innovative solutions more effectively
to users and networks. To demonstrate this approach,
we utilised the Dynamic Sustainability Framework,
applying it to epidemic modelling scenarios for bot
propagation. We suggested measurement techniques,
highlighted key constructs, and discussed the evalua-
tive process. IS deals with the impact and implemen-
tation of an intervention, allowing us to consider how
we develop multi-faceted approaches, how/where we
deliver them, what meaningful impact they have, and
why they may be lacking. This is vital for improving
and sustaining the health of networks. We hope that
this work contributes towards a discussion about how
we deliver new solutions and how TR may play a role
in this.
REFERENCES
Bassam, S. F. and Deborah, L. F. (2010). Cyber Security
Framework for Enterprise System Development: En-
hancing Domain Security through ESM. In Military
Communications Conference, 2010-MILCOM 2010,
pages 924–929. IEEE.
Bauer, M. S., Damschroder, L., Hagedorn, H., Smith, J., and
Kilbourne, A. M. (2015). An Introduction to Imple-
mentation Science for the Non-Specialist. BMC Psy-
chology, 3(1):32.
Brauer, F. (2008). Compartmental Models in Epidemiol-
ogy. In Mathematical Epidemiology, pages 19–79.
Springer.
Chambers, D. A., Glasgow, R. E., and Stange, K. C. (2013).
The Dynamic Sustainability Framework: Addressing
the Paradox of Sustainment Amid Ongoing Change.
Implementation Science, 8(1):117.
Dagon, D., Zou, C. C., and Lee, W. (2006). Modeling Bot-
net Propagation Using Time Zones. In NDSS, vol-
ume 6, pages 2–13.
Dedeke, A. (2017). Cybersecurity Framework Adoption:
Using Capability Levels for Implementation Tiers and
Profiles. IEEE Security Privacy, 15(5):47–54.
Galea, S. (2013). An Argument for a Consequentialist
Epidemiology. American Journal of Epidemiology,
178(8):1185–1191.
Khalil, H. (2016). Knowledge Translation and Implemen-
tation Science: What is the Difference.
Khouzani, M., Altman, E., and Sarkar, S. (2012). Optimal
Quarantining of Wireless Malware through Reception
Gain Control. IEEE Transactions on Automatic Con-
trol, 57(1):49–61.
Lu, Z., Wang, W., and Wang, C. (2016). On the Evolu-
tion and Impact of Mobile Botnets in Wireless Net-
works. IEEE Transactions on Mobile Computing,
15(9):2304–2316.
NIST (2014). Framework for Improving Critical Infrastruc-
ture Cybersecurity. Last accessed 5th January 2018.
Rubio, D. M., Schoenbaum, E. E., Lee, L. S., Schteingart,
D. E., Marantz, P. R., Anderson, K. E., Platt, L. D.,
Baez, A., and Esposito, K. (2010). Defining Trans-
lational Research: Implications for Training. Aca-
demic Medicine: Journal of the Association of Ameri-
can Medical Colleges, 85(3):470.
Sanzgiri, A., Joyce, J., and Upadhyaya, S. (2012). The Early
(Tweet-ing) Bird Spreads the Worm: An Assessment
of Twitter for Malware Propagation. Procedia Com-
puter Science, 10:705–712.
Woolf, S. H. (2008). The Meaning of Translational Re-
search and Why it Matters. Jama, 299(2):211–213.
Yan, G., Chen, G., Eidenbenz, S., and Li, N. (2011). Mal-
ware Propagation in Online Social Networks: Nature,
Dynamics, and Defense Implications. In Proceedings
of the 6th ACM Symposium on Information, Computer
and Communications Security, pages 196–206. ACM.
Yong, W., Tefera, S. H., and Beshah, Y. K. (2012). Under-
standing Botnet: From Mathematical Modelling to In-
tegrated Detection and Mitigation Framework. In Soft-
ware Engineering, Artificial Intelligence, Networking
and Parallel & Distributed Computing (SNPD), 2012
13th ACIS International Conference on, pages 63–70.
IEEE.
Evaluating the Provision of Botnet Defences using Translational Research Concepts
519