Adopting an Agent and Event Driven Approach for Enabling Mutual
Auditability and Security Transparency in Cloud based Services
Mou
ssa Ouedraogo¹, Eric Dubois¹, Djamel Khadraoui¹, Sebastien Poggi² and Benoit Chenal²
¹Luxembourg Institute of Science and Technology, 5 Avenue des hauts Fourneaux, L4362 Esch/Alzette, Luxembourg,
Luxembourg
²Victor Buck Services S.A, L-8308 Capellen, Luxembourg, Luxembourg
Keywords: Cloud, Security Transparency, Mutual Auditability, Monitoring, Event Specification and Detection.
Abstract: We propose an event-driven approach for the automated audit of cloud based services security. The pro-
posed approach is a solution to two of the intrinsic security issues of cloud based services, notably the need
of security transparency and mutual auditability amongst the stakeholders. We leverage a logic based event
specification language to represent patterns of events which occurrence can be evidence of security anomaly
or breach or simply a sign of a nefarious use of the cloud infrastructure by some of its users. The use of ded-
icated algorithms for the detection of composite events coalesced with the definition of primitive events
structure based on XCCDF format ensures the reuse and interoperability with security audit tools based on
the Security Content and Automation Protocol-SCAP. The implementation and application of the approach
on a cloud service dealing with electronic archiving have demonstrated its feasibility and viability.
1 INTRODUCTION
For most businesses and individuals, Cloud based
services are the alternative to achieving cost-
efficiency in the provisioning and consumption of
services. However companies dealing with security
and/or privacy critical data, have often shown some
reluctance to fully embrace the trend, even if there is
evidence that the trend is starting to sift at least for
the banking and financial sector (http://
www.businesscloudnews.com/2014/06/02/cloud-in-
financial-services-what-is-it-not-good-for/). Several
factors could explain such an attitude towards the
cloud: In the cloud, the data and the mechanisms
necessary for its processing may reside in the pro-
vider’s premises. This leads to some devolution of
security matters about such data and processes to the
cloud provider whose capability and/or due dili-
gence to deal with the security issues may be mis-
trusted or simply feeble. The uncertainty on the ac-
tual location of the data is also exacerbated by the
complexity of the chain of provider-consumer. In
fact, although a CSP may be registered in a given
country, the chain of provider-consumer may be
such that the actual data centre used by the CSP is
located elsewhere. Given the stored information may
be subject to the legislation of the country where it is
stored physically, this may also pose serious privacy
management challenges. In fact there may be ambi-
guity in understanding which regulation applies for a
data about a third country citizen (which should
normally be subjected to national regulation) but
stored in another country, where regulation towards
privacy may be well different. The multi-tenancy
aspect that is most often used to characterize cloud
computing also introduces a new risk unique to
cloud services, the possibility of attacks from other
consumers, who may be competitors or simply
hackers, co-located on the same infrastructure, e.g.,
servers, hard disks, virtual machines. This is well
exemplified by “Amazon Zeus botnet” incident in-
volving Amazon EC2’s infrastructure (McAfee and
Guardian Analitics, 2012), whereby cybercriminals,
by initially hacking into a service hosted by Amazon
cloud infrastructure, were able to install command-
and-controls infrastructure with the aim to infect
client computers and steal their banking credentials.
This incident is a reminder that the security of the
cloud service is only as good as at its weakest link
given that a vulnerability at a tenant application may
result in the jeopardy of the whole service. This
status quo calls for techniques that help to foster
more security assurance in the cloud realm. Security
assurance being the ground for confidence that secu-
rity deployed and/or managed by a third party is
correctly implemented and also effective against the
565
Ouedraogo M., Dubois E., Khadraoui D., Poggi S. and Chenal B..
Adopting an Agent and Event Driven Approach for Enabling Mutual Auditability and Security Transparency in Cloud based Services.
DOI: 10.5220/0005496205650572
In Proceedings of the 5th International Conference on Cloud Computing and Services Science (CLOSER-2015), pages 565-572
ISBN: 978-989-758-104-5
Copyright
c
2015 SCITEPRESS (Science and Technology Publications, Lda.)
risks (Ouedraogo et al., 2012). In third party services
such as the cloud, security assurance can be practi-
cally met by probing the security of the CSP through
audits and by gaining more visibility on its security
policy and operation through security transparency
mechanism (Winkler, 2011). Consequently, achiev-
ing a wider adoption of cloud based services would
depend on how effective issues related to mutual
auditability and security transparency can be ad-
dressed (Chen, 2010; Ouedraogo et al., 2013; Nuñez
et al., 2013; Sunyaev and Schneider, 2013). Tech-
nics and approaches tailored in that vein of idea
should enable the Cloud Service Consumer (CSC),
provided the existence of contractual clauses with
the Cloud Service Provider (CSP), to gather evi-
dence that corroborate or challenge compliance,
performance and security claim made by the CSP,
while at the same time enabling the latter to monitor
the activity and traffic of the users to ensure no
abuse and nefarious use of the cloud is made.
This paper’s contribution can be summarized as
an effort to leverage Event-driven computing (Luck-
ham, 2005; Etzion and Niblett, 2010) and Multi
agent systems-MAS- (Ganzha and Paprzycki, 2014)
to foster more security transparency and enable mu-
tual-auditability in a cloud setting. To achieve this,
we resort to a tree based specification of security
events of interest by the CSC and CSP while within
the infrastructure, software agents are generated for
capturing such events in case they materialize. We
amalgamate real time security related event detec-
tion and logic-based rules for empowering both the
CSC and CSP, with effective means of depicting and
promptly detecting anomalies and security or QoS
breaches. An event is here considered as a happen-
ing of interest (related to security or quality of ser-
vice procurement) to the CSP or CSC.
The paper is organized as follows: Section 2
analyses the related work. Section 3 provides a de-
scription of the adopted architecture. In Section 4
we specify monitor-able event using a logic based
language. Section 5 shows how audits and transpar-
ency are enforced, while Section 6 presents an appli-
cation case. Section 7 provides some concluding
remarks.
2 RELATED WORKS
Initiatives purporting to address the issue of mutual
trust and transparency in the cloud have mainly re-
volved around the topic of audit, Virtual machine
introspection and Service level agreement. Audits
standards including SSAE16 (www.ssae16.com),
and its international version ISAE3402 (http://
isae3402.com/) rely in a large part on the words and
assessment of the CSP, information that cannot be
guaranteed to be immune from bias. Recent efforts
in cloud audits have leaned towards automation. For
instance, the CSA cloud-audit (http://cloudaudit.org/
CloudAudit/Home.html) pur-ports the automation of
standard audit and related assurance and compliance
effort by providing a controlled set of interfaces to
allow CSCs or their representatives to assess their
services. Dolitszcher et al. (2013) propose a cloud
audit methodology based on the usage of MAS for
conducting the audit of virtual machines dynamical-
ly allocated to clients to account for changes within
the cloud infrastructure.
Rak et al. (2011) adopts APIs derived from the
mOSAIC project (http://www.mosaic-project.eu/) to
build up an SLA-oriented cloud application that
enables the management of security features related
to user authentication and authorization. An exten-
sion of the work of Rak et al. can be found through
the EU FP7 project Specs aiming to deliver a plat-
form for providing a security services based on SLA
management. The SLA monitoring in SLA@SOI
relies on EVEREST+ (Lorenzoli and Spanoudakis,
2010), which is a general-purpose engine for moni-
toring the behavioural and quality properties of dis-
tributed systems based on events captured from them
during the operation of these systems at runtime.
The major problem with the adoption of SLA man-
agement as a means to enhance security transparen-
cy is primarily on its practicality. Indeed the aca-
demic notion of SLA appears to be far more exten-
sive than it is in reality. In the context of this work,
the authors have approached a number of CSPs in
Luxembourg with the aim to get a glimpse on the set
of items that were part of their SLA. Most often,
such documents were restricted to the sole aspects of
allocated bandwidth, storage capacity, etc..; while
the only security aspect included was related to ser-
vice availability. Clearly, the items included in those
specifications were those the companies were confi-
dent they could deliver on. Their argument on the
most pressing and challenging issues such as securi-
ty was that stringent mechanisms were in place for
its guarantee as evidenced by their certifications.
Unlike the existing initiatives, our approach lev-
erages events processing to enable mutual audit
between the CSC and CSP. While existing commer-
cial and open source solutions for event analytics
such as Splunk (Carasso, 2012), Arcsight (http://
www.arcsight.net/) and Graylog2 (https://
www.graylog.org/graylog2-v0-92/) are based on log
analysis, our initiative is based on near-real time
CLOSER2015-5thInternationalConferenceonCloudComputingandServicesScience
566
Figure 1: Architecture of the event specification monitoring for transparency and mutual auditability.
detection of primitive events followed by a reason-
ing on whether the specified composite event has
materialized. Thus, allowing the concerned
CSC/CSP to promptly take mitigating actions. Also,
provided the existence of a contractual agreement,
both the CSP and CSC can specify and launch a
monitoring of the security of the other.
Additionally our approach is that the event driven
specifications provides a good expressivity for cap-
turing events of interest emanating from an SLA, an
internal policy or external regulations while allow-
ing the reuse of existing security management tools.
3 HIGH LEVEL ARCHITECTURE
OF THE APPROACH
Our event driven approach and tool is made of three
main components as depicted in Figure 1.
The first component is a case tool or Event de-
signer that offers a graphical design interface to the
cloud stakeholders, for specifying patterns of events
that are of interest. Upon the design of the events
composites that could result from combining differ-
ent patterns to monitor, the stakeholder provides
technical details for each event using an event speci-
fication format of choice. In the context of our work,
the Extensible Configuration Checklist Description
(XCCDF) format (Waltermire et al., 2011) is adopt-
ed as further elaborated in Section 4.1. The case tool
also serves as a dashboard for the visualization of
the events status once the monitoring is triggered.
The second component relates to the elaboration
of a multi-agent system embedded within the cloud
infrastructure (CI) with the purpose of detecting
each primitive events specified within the Event
Designer console. The definition and management of
the detection agents are performed using JADE plat-
form (Bellifemine et al., 2008). The peculiarity of
the agent structure and organization is adopted from
the work of Ouedraogo et al. (2014).
The third component is the event processing lay-
er. Individual atomic events captured from the cloud
infrastructure are sent by specialized agents to the
reasoning layer where dedicated algorithms detailed
in Section 5 will be resorted for informing the stake-
holder when a specified pattern has materialized.
In the following we further elaborate on how
those three components play a role in practically
delivering mutual audit and help foster better trans-
parency.
4 SPECIFICATION OF
MONITORABLE EVENTS
The decision to adopt an event driven approach to
audit and monitor the security in the cloud is under-
score by the argument that events provide a power-
ful construct to capture current state of a system and
deviations from expectation and to predict future
security or QoS related issues ( Luckham, 2005;
Etzion and Niblett, 2010). Additionally a well-
defined architecture can support event based moni-
Reasoning Layer
Cloud Infrastructure &
with Detection agents
CSP
Event Designer
and Visualization
CSC
AdoptinganAgentandEventDrivenApproachforEnablingMutualAuditabilityandSecurityTransparencyinCloud
basedServices
567
toring in ensuring the prompt dissemination of its
occurrence to the interested parties who would make
judgment on the course of action to adopt. Amongst
others, it may be a way to hold cloud providers ac-
countable for a security breach that may has
stemmed from a lax in their security; a breach of
SLA or other escrows between the two parties. The
set of patterns and the detection algorithms associat-
ed could also constitute a powerful tool for a cloud
provider concerned with activities of its clients.
A prerequisite for effective event patterns detec-
tion is the definition of a clear event structure cou-
pled with the adoption of a pattern specification
language that is expressive enough to capture the
realm of events of interest and their propensity. Only
after that one can begin to implement the required
strategy for the ensuing detection. This section pro-
vides an insight into our event based approach.
4.1 A Primitive Event Structure to
Ensure Re-Use of Existing Security
Tools
Commonly an event is defined as an occurrence of
interest within a system or domain (Etzion and Nib-
lett, 2010). Subsequently, dealing with events could
purport the monitoring of a system or process with
the intent to flag exceptional or anomalous behav-
iors. Alternatively, event analysis could be the base-
line for (i) predicting a major event before they actu-
al takes place as in fraud detection application, fi-
nancial market trends and natural disaster; (ii) diag-
nosing a problem based on deductive reasoning after
the observation of symptomatic events. While most
event structure includes header information that
provides meta-information about the event (identifi-
cation number, occurrence time, description, catego-
ry, etc...), attributes related to the event payload is
intrinsically linked to the intent sought for their pro-
cessing. For instance, for the description of an event
structure pertinent to a credit card fraud, the geo-
graphical locations where individual purchase takes
place and the amount of money involved are very
salient information to capture. Owing to the fact that
this audit emphasis on anomalies related to security,
the idea was then to adopt an event structure that
could allow the re-use of existing security audits
tools given the area of network and system audits is
already beaming with a plethora of tools. The adop-
tion of an Extensible Configuration Checklist De-
scription Format (XCCDF) like format as a baseline
for the primitive events structure was thus to ensure
reuse and interoperability with Security Content and
Automation Protocol (SCAP) tools. XCCDF is an
XML based format used to specify security check-
lists and benchmarks amongst others. The overarch-
ing purpose of the format is to provide a uniform
expression of security checklists, benchmarks, and
other configuration guidance, and thereby foster
more widespread application of good security prac-
tices. With analogy to the XCCDF format, we speci-
fy an event with the following attributes:
<Name>..<\Name>
<Identification>..<\Identification>
<Description>..<\Description>
<Category>.. <\Category>
<Time stamp>.. <\Time stamp>
<Value>..<\Value>
<Frequency>..<\Frequency>
<Rule>..<\Rule>
<Probe>..<\Probe>
In the context of this work, the most relevant attrib-
utes associated to a primitive event include the rule
attribute which is the underlining policy based on
which the associated probe (either a SCAP tool or an
in-house program) could interpret and carry out the
specificities of the rule, leading to the detection of
the primitive event. Furthermore, the rule attribute
encompasses and provides reasoning about the ex-
pected and exceptional behaviours and states. The
field associated to a rule take as input a path leading
to a file, thus allowing one to define a comprehen-
sive set of policy that should drive the detection of
primitive events of interest. A rule could be speci-
fied in a logical language such as Etalis (Anicic et
al., 2012) and Drools fusion (http://
www.drools.org/) or programing language including
Python as in the case of our implementation. The
category attribute is a field we added, for allowing
the user to systematically classify events based on
their typology and/or interest for the stakeholders.
Secondary event attributes such as name, identifica-
tion, description, timestamp, frequency, denote re-
spectively, the given name unique identifier of the
event, a succinct description of it, the time at which
the event was created, the frequency at which the
tool should be probing the event.
4.2 Event Patterns Specification
We hereafter use the term event pattern to refer to
any composite event whereby primitive events
(leaves of the tree) are associated through a logical
operator or connector. An event tree can thus be a
simple event patterns or a combination of patterns of
different semantic leading to a much complex event
specification. Adopting a tree based representation
also allows leveraging relevant graph theories for the
efficient processing of the event nodes. Our event-
based cloud audit and monitoring is based on event
CLOSER2015-5thInternationalConferenceonCloudComputingandServicesScience
568
patterns specification using the Yet Another Lan-
guage for Event Specification or YALES logic
(Zhang and Unger, 1996). The choice of YALES
lies on its expressiveness but also to the fact that it
allows to capture synchronous and asynchronous
events and also provides event counter and calendar
constructs. YALES distinguishes three types of
events: temporal events denoting the explicit time
instants in real life and can be deemed relative tem-
poral event when the offset is equal to span. Calen-
dar events express the periodical activities that occur
regularly or irregularly many times in terms of real
life time measurements and primitive events. A
primitive event is considered as an explicit event
taking place at a discrete or during a time span.
More precisely, our framework and tool allows both
the cloud client and provider to declare and detect
the following event patterns or event composite
(note that this list of pattern is not exhaustive and
new one can be added for detection purpose):
Disjunction of Events, E
1 E2 ... En or any
Pattern: This pattern of events, occurs when one or
more of primitive events E1,..,En occurs. The detec-
tion algorithm for the ANY pattern is later provided.
Conjunction of Events, A (m, {E1, E2 ... En})|
Part of: This pattern of event occurs if all the con-
stituent events occur. With a slight extension and
with the notation of A (m, {E1, E2... En}) where m
< n, we can express the idea that if at least m out of
n events occur, the event pattern has happened.
Sequences of Events, E1; E2; ..; En | All in Or-
der: A sequence of events pattern expresses the
requirement that occurrence of composite events be
strictly in order in time, i.e. no adjacent events that
occur at the same time are counted.
Event Counter, C (E, n^ | n | n+ | n-): | Over-
time, Equals, More, Less: A way to tell how many
times an event has occurred is useful. The event
counter as an event pattern is designed to provide
this kind of mechanism. Event counter, C (E, n^),
will occur upon every nth occurrence of E. In this
case C (E, n^) may be triggered more than once; C
(E, n) is validated at when at the nth occurrence of
E. Thus C (E, n) is only triggered once. C (E, n+) is
validated when E has occurred not less than n times;
while C (E, n-) will be flagged when E has occurred
less than n times but at least once.
Moving Window, W (n, E, span): A moving
window uses a moving interval with a fixed span to
provide aggregate event information. Here, W (n, E,
span) is used to mean that if there are more than n of
E occurrences during a time period, then the compo-
site event should be raised.
Put the Occurrence of an Event in Context of a
Sequence of Other Events E: C | [E1, E2] and E
IN C | [E1, E2]: Events in periods or intervening
events are those that occur in a period marked by
two reference events E1and E2. Two alternative
cases can be considered: The first supposes a left-
closed and right-open interval that we refer to as
BETWEEN event pattern; and the second one con-
sidering a left-closed and right-closed interval re-
ferred to as IN event pattern. While both detection
patterns BETWEEN and IN would require the event
of interest to occur after the initiating event (E1), the
IN pattern is detected only after the right-end event
has occurred as opposed to the BETWEEN event.
5 FROM EVENTS
SPECIFICATION TO AUDIT
The previous section provided the foundation for
specifying the patterns of events relevant to the se-
curity audit. In this section, we depict how in prac-
tice patterns are used and analyzed to support mutual
auditability and increase cloud transparency. The
mutual audit process and the ensuing increase in
transparency between the CSC and the CSP is sup-
ported by two main steps once the composite events
have been specified: the generation and triggering of
software agents for conducting detecting the primi-
tive events and the reasoning on whether an overall
event pattern of interest has taken place. This will
ultimately lead to the generation of reports and alerts
on the dashboard of the cloud stakeholder of interest.
5.1 Generation and Triggering of
Agents
To conduct the audits, an organization of software
agents is used. The first type of agent or probe
agents, purport to conduct the detection of the primi-
tive events within the pattern specified by the cloud
stakeholder. The second type of agent is a single
agent in some cases, referred to as Event receiver
which role is to filter and aggregate the set of inputs
that arrive from the probe agents before they are
passed to the reasoning engine. In order to ensure
the intrinsic link between a primitive event and a
probe agent, we adopt the following reference for-
mat for agent during their generation:
Agent<EventName>@<Domain>, where domain
refers to the name given to the audit platform; and
eventName- a unique name given to the event in the
event structure provided in Section 3.1. In practice,
AdoptinganAgentandEventDrivenApproachforEnablingMutualAuditabilityandSecurityTransparencyinCloud
basedServices
569
the probes agents actually carrying out the instruc-
tion within the Rule field of the primitive event, thus
triggering an existing security tool or launching a
pseudo code before collecting the results of the
check. The creation and management of the agents
can be done through a MAS platform such as JADE.
5.2 Reasoning and Detecting a Pattern
Primitive events alone may not always be significant
to portray the emergency of a situation. In contrast,
considering a pool of events from the same or differ-
ent sources within the infrastructure of the CSC/CSP
may reveal a pattern that relate to an anomaly or
impending risk for the service stakeholder. As men-
tioned in the previous section, since we have adopt-
ed a tree based representation and validation of
event patterns, efficiently detecting a pattern of
events depends on how the tree is processed, and
individual events are handled upon their detection.
With this in mind, we have tailored for each of the
event pattern specified in Section 3.2, a dedicated
algorithm that is resorted by the reasoning engine to
determine whether the composite event has material-
ized. Owing to page limitation, we only provide a
description of some amongst them:
Any: this operator has a list of events and is validat-
ed if one of those events occurs. If the operator
above this operator is a count operator, then
this operator will reset only the event that validated
it. This ensures that any other subsequent event in
the list that was partially validated, will still be ac-
counted for.
Algorithm:
boolean valid = false
for each event in list {
if event is validated {
add event to happenedList
valid = true
}
}
return valid
Part Of: This operator has a list of events and a
trigger number. It is validated when the number of
events validated in the list is equal or above to the
trigger number. If the operator above this operator is
a count operator, this operator will reset only the
events of the list that are validated. Therefore if
another event in the list was partially validated, it
will not lose its current state.
Algorithm :
clearhappenedList
for each event in list {
if event is valdiated {
add event to happenedList
}
}
if happenedList size >= trigger {
return true
}
return false
5.3 Audit Reporting
Upon the processing of the events by the reasoning
engine, it sends the result to the Report Generator
which subsequently displays it at the HMI or Dash-
board. Given the adoption of a tree based specifica-
tion of the composite events to monitor for, the
Event designer could also serve the purpose of visu-
al dashboard whereby events detected by the agents
get automatically highlighted in a different colour.
For further details information on the detected
events, another dashboard could be added, giving
details on the event’s name, primary identifier , date
of creation, the probe agent tasked with detecting it,
and its status (for instance Occur or not happened).
Such information could then allow the CSP and CSC
to respectively, to detect any nefarious and mali-
cious use of its service by a CSC, detect any security
anomaly within their virtual machines and/or held
the CSP accountable for a breach of contractual
agreement related to security and quality of service.
6 APPLICATION TO A REAL
SCENARIO
The scenario described thereafter is a real use case
we have worked on though the names of the compa-
nies involved have been altered.
L-BANK which is actually affiliated to an inter-
national Bank, has decided to use the service of
D.CLOUD owing to the fact that the local branch
was closing down. As information related to the
bank customers could not be moved beyond the
boundary of the country, it was therefore imperative
to found a reliable service that could carry out the
archiving in accordance with the legal framework.
D.CLOUD was chosen primarily for the fact that
archiving was the core of its business and also be-
cause of the security reliability (supported by certifi-
cations). This means a considerable amount of ar-
chives (bank customers’ information) were ready to
be transferred to D.CLOUD through SFTP connec-
tion mode for fast integration. L-BANK was then
given an access to D.CLOUD’s standard web portal.
D-CLOUD could create new archives, search, re-
store, and consult individual archives. This portal is
accessible directly over Internet (HTTPS) or through
a VPN. At the end of the retention period, the
CLOSER2015-5thInternationalConferenceonCloudComputingandServicesScience
570
Figure 2: Launch of a JADE based agents and visualization of archive’s related events.
documents are available for deletion or archiving.
From L-BANK’s view point the following events
were identified as being relevant for monitoring:
-Confidentiality of the ANY (C1, C2) where
C1: An attacker or the cloud provider staff may ac-
cess. the archive stored on the cloud provider server
without authorisation.
C2: An error in storage or a bad virtual separation
may cause a breach of confidentiality.
- Integrity of the Archives ANY (I1, I2, I3)
I1: Unauthorised modification by hackers may occur
when the archive is stored on the CSP’s server.
I2: A loss of integrity of the archive due to a corrup-
tion of data may occur when the cloud client con-
sults the archive through the archive portal.
I3: In the documents management configuration,
when the client consults the archive, an error may
cause a loss of integrity or a deletion of the archive.
-Availability of the Archives ANY (A1, A2) where
A1:The platform may be unavailable due to a traffic
overloading or a denial of service attack.
A2: Transmission and communication errors may
occur when the archive is send to the cloud client.
The D-CLOUD in turn was concerned with un-
warranted modification and activities from L-BANK
with consequences on its services:
ANY (M1, M2) where:
M1:The cloud client may intentionally or not upload
a malicious file as a archive.
M2: an error in the system or archive maintenance
may result in unwanted changes.
Noteworthy, the specifcation of composite events
combining events from the three groups is possible
as it was the case during the application to the use
case (Figure 1-2). Similarly most of the events above
listed can be further specified as composite events of
their own, but due to page restriction such an exr-
cicse will not be conducted in this paper.
7 CONCLUDING REMARKS
This paper has reported on an initiative for address-
ing security transparency and allowing mutual au-
ditability within a cloud realm. Given that the use of
raw logic based language may prove a challenge for
some stakeholders wishing to engage in the system-
atic audit and monitoring of security and QoS relat-
ed parameters, the choice of a tree based representa-
tion was made. Another key aspect of the approach
is the adoption of the XCCDF format, which enables
the reuse of existing security management tools.
AdoptinganAgentandEventDrivenApproachforEnablingMutualAuditabilityandSecurityTransparencyinCloud
basedServices
571
The initial prototype based on the concepts and algo-
rithms presented has been validate on an electronic
archiving platform with the event specification and
detection console allocated to a dedicated virtual
machine, while the multi agent system platform
JADE has been adopted for the specification and
management of agent entrusted with the role of de-
tecting primitive events as can be seen in Figure 2.
NAGIOS plugins (
Pervilä, 2007) along with other
tailored programs where developed for the detection
of primitive events within the Infrastructure.
The initial results were very encouraging as most
of the security events of concerns provided by the
SaaS provider and consumer and specified using the
Event designer were detecting, by simulating altera-
tions and attacks targeting the archived files. Fur-
thermore, the capacity of the VM required for host-
ing the whole application (Event Designer and mul-
ti-agent detection platform) was confine to a 2 Go of
RAM and in single CPU. Nonetheless, further appli-
cations are envisaged for better appraising the effect
of deploying simultaneously a multitude of agents
for detecting and reporting events of interest.
AKNOWLEDGEMENTS
This work has been conducted in the context of the
SAINTS project, financed by the national fund of
research of the Grand Duchy of Luxembourg (FNR)
under grant number C12/IS/3988336. The authors
also thanks Maimouna Seck and Charles Hubert
Duthilleux for their work on implementing the tool.
REFERENCES
Anicic D., Rudolph S., Fodor P., Stojanovic N.: Stream
reasoning and complex event processing in ETALIS.
Semantic Web 3(4): 397-407 (2012).
Bellifemine F., Caire G, Poggi A., Rimassa G. 2008
JADE: A software framework for developing multi-
agent applications. Lessons learned. Information &
Software Technology 50(1-2): 10-21.
Carasso D. (2012) Exploring Splunk, CITO Research,
New York.
Chen Y, Paxson V, Katz RH (2010) What’s New About
Cloud Computing Security? Report EECS Depart-
ment, University of California, Berkeley,
http://www.eecs.berkeley.edu/Pubs/TechRpts
/2010/EECS-2010-5.html.
Dölitzscher F., Knahl M., Reich C., Clarke N.L. 2013
Anomaly Detection in IaaS Clouds. In proceedings of
CloudCom (1) 387-394.
Etzion O., Niblett P. 2010. Event Processing in Action.
Manning Publications Company 2010, ISBN 978-1-
935182-21-4, pp. I-XXIV, 1-360.
Lorenzoli D., Spanoudakis G. 2010 EVEREST+: Runtime
SLA Violations Prediction: In: Proceedings of the 5th
Middleware for Service-oriented Computing Work-
shop, ACM.
Luckham D. C. (2005) The power of events - an introduc-
tion to complex event processing in distributed enter-
prise systems. ACM 2005, ISBN 978-0-201-72789-0,
pp. I-XIX, 1-376.
Ganzha M, Paprzycki M. (2014): Agent-oriented compu-
ting for distributed systems and networks. J. Network
and Computer Applications 37: 45-46 (2014). McAfee
and Guardian Analytics. 2012. Dissecting. Operation
High Roller. Accessed 10 December 2014. From:
http://www.mcafee.com/us/resources/reports/rp.operat
ion-high-roller.pdf.
Nuñez D., Fernandez – Gago C., Pearson S., Felici M.
2013 A Metamodel for Measuring Accountability At-
tributes in the Cloud. In: Proceedings of the 2013
IEEE International Conference on Cloud Computing
Technology and Science (CloudCom 2013), IEEE.
Ouedraogo M., Khadraoui D., Mouratidis, H. and Dubois
E. (2012): Appraisal and reporting of security assur-
ance at operational systems level. Journal of Systems
and Software 85(1): 193-208 (2012).
Ouedraogo M, Mouratidis M (2013) Selecting a cloud
service provider in the age of cybercrime, Computers
& Security, vol.38, pp.3-13 Special issue on Cyber-
crime in the Digital Economy, Elsevier.
Ouedraogo M., Kuo C.T, Tjoa S., Preston D, Dubois E.,
Simões P., Cruz T.: Keeping an Eye on Your Security
Through Assurance Indicators. In proceedings of
SECRYPT 2014: 476-483.
Pervilä, M.A., 2007. Using Nagios to monitor faults in a
self-healing environment. In:Seminar on Self-Healing
Systems. University of Helsinki.
Rak M, Liccardo L, Aversa R 2011. A SLA-based inter-
face for security management in cloud and GRID inte-
grations. In: Proceedings of the 7
th
International. Con-
ference on Information Assurance and Security (IAS),
pp.378-383, IEEE.
Robert J. Zhang, Elizabeth A. Unger (1996) Event Speci-
fication and Detection Technical report TR CS-96-8,
1996, Kansas State University.
Sunyaev A., Schneider S. 2013. Cloud services. certifica-
tion Communication of the ACM 56(2): 33-36, ACM
digital Library.
Waltermire D., Schmidt, C., Scarfone K.
Winkler V. (2011) Securing the cloud- cloud computer.
security techniques and tactics. Syngress.
Ziring N. 2012. Specification for the Extensible Configu-
ration Checklist Description Format (XCCDF) Ver-
sion 1.2, NIST Interagency Report 7275Revision 4,
National Institute of Standards and Technology
Gaithersburg, MD 20899-89.
CLOSER2015-5thInternationalConferenceonCloudComputingandServicesScience
572