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

net/publication/255708329

An analysis of the cloud computing security problem

Conference Paper · January 2010

CITATIONS READS

271 3,652

3 authors:

Mohamed Almorsy John Grundy

Swinburne University of Technology Monash University (Australia)
26 PUBLICATIONS   900 CITATIONS    616 PUBLICATIONS   8,235 CITATIONS   

SEE PROFILE SEE PROFILE

Ingo Mueller
Monash University (Australia)
24 PUBLICATIONS   553 CITATIONS   

SEE PROFILE

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

Software for health and social services View project

Software Process and Product Improvement (SPPI) View project

All content following this page was uploaded by John Grundy on 26 May 2014.

The user has requested enhancement of the downloaded file.

Swinburne Research Bank
http://researchbank.swinburne.edu.au

Al Morsy, M., Grundy, J. & Mueller, I. (2010). An analysis of the cloud computing
security problem.

Paper presented at the 17th Asia-Pacific Software Engineering Conference

(APSEC 2010) Cloud Workshop, Sydney, Australia, 30 November-03 December
2010.

Copyright © 2010.

This is the author’s version of the work, posted here with the permission of the author
for your personal use.

Swinburne University of Technology | CRICOS Provider 00111D | swinburne.edu.au

 In Proceedings of APSEC 2010 Cloud Workshop, Sydney, Australia, 30th Nov 2010.

An Analysis of The Cloud Computing Security Problem

Mohamed Al Morsy, John Grundy and Ingo Müller
Computer Science & Software Engineering, Faculty of Information & Communication Technologies
Swinburne University of Technology, Hawthorn, Victoria, Australia
{malmorsy, jgrundy, imueller}@ swin.edu.au

Abstract — Cloud computing is a new computational paradigm attackers’ interest in finding existing vulnerabilities in the
that offers an innovative business model for organizations to model.
adopt IT without upfront investment. Despite the potential gains Despite the potential benefits and revenues that could be
achieved from the cloud computing, the model security is still gained from the cloud computing model, the model still has a
questionable which impacts the cloud model adoption. The
lot of open issues that impact the model creditability and
security problem becomes more complicated under the cloud
model as new dimensions have entered into the problem scope pervasiveness. Vendor lock-in, multi-tenancy and isolation,
related to the model architecture, multi-tenancy, elasticity, and data management, service portability, elasticity engines, SLA
layers dependency stack. In this paper we introduce a detailed management, and cloud security are well known open research
analysis of the cloud security problem. We investigated the problems in the cloud computing model.
problem from the cloud architecture perspective, the cloud From the cloud consumers’ perspective, security is the major
offered characteristics perspective, the cloud stakeholders’ concern that hampers the adoption of the cloud computing
perspective, and the cloud service delivery models perspective. model [3] because:
Based on this analysis we derive a detailed specification of the
• Enterprises outsource security management to a third party
cloud security problem and key features that should be covered by
any proposed security solution.
that hosts their IT assets (loss of control).
• Co-existence of assets of different tenants in the same
Keywords: cloud computing; cloud computing security; cloud location and using the same instance of the service while
computing security management. being unaware of the strength of security controls used.
I. INTRODUCTION • The lack of security guarantees in the SLAs between the
Cloud computing provides the next generation of internet- cloud consumers and the cloud providers.
based, highly scalable distributed computing systems in which • Hosting this set of valuable assets on publicly available
computational resources are offered 'as a service'. The most infrastructure increases the probability of attacks.
widely used definition of the cloud computing model is From the cloud providers’ perspective, security requires a lot of
introduced by NIST [1] as “a model for enabling convenient, expenditures (security solutions’ licenses), resources (security
on-demand network access to a shared pool of configurable is a resource consuming task), and is a difficult problem to
computing resources (e.g., networks, servers, storage, master (as we discuss later). But skipping security from the
applications, and services) that can be rapidly provisioned and cloud computing model roadmap will violate the expected
released with minimal management effort or service provider revenues as explained above. So cloud providers have to
interaction.”. Multi-tenancy and elasticity are two key understand consumers’ concerns and seek out new security
characteristics of the cloud model. Multi-Tenancy enables solutions that resolve such concerns.
sharing the same service instance among different tenants. In this paper we analyze existing challenges and issues
Elasticity enables scaling up and down resources allocated to a involved in the cloud computing security problem. We group
service based on the current service demands. Both these issues into architecture-related issues, service delivery
characteristics focus on improving resource utilization, cost model-related issues, cloud characteristic-related issues, and
and service availability. cloud stakeholder-related issues. Our objective is to identify
The cloud model has motivated industry and academia to the weak points in the cloud model. We present a detailed
adopt cloud computing to host a wide spectrum of applications analysis for each weakness to highlight their root causes. This
ranging from high computationally intensive applications down will help cloud providers and security vendors to have a better
to light weight services. The model is also well-suited for small understanding of the problem. It also helps researchers being
and medium businesses because it helps adopting IT without aware of the existing problem dimensions and gaps.
upfront investments in infrastructure, software licenses and Our paper is organized as follows. In section II, we explore
other relevant requirements. Moreover, Governments become previous efforts in defining cloud security problems and
more interested in the possibilities of using cloud computing to challenges. Sections III to VII explore the cloud computing
reduce IT costs and increase capabilities and reachability of security problem from different perspectives. Section VIII
their delivered services. discusses the key security enablers in the cloud model. Section
According to a Gartner survey [2] on cloud computing IX summarizes our conclusions and what we believe are the
revenues, the cloud market was worth USD 58.6B in 2009 , is key dimensions that should be covered by any cloud security
expected to be USD 68B in 2010 and will reach USD 148B by solution. Finally, in section X we discuss the future work
2014. These revenues imply that cloud computing is a focusing on one of the discussed security enablers (cloud
promising platform. On the other hand, it increases the security management).
 II.LITERATURE REVIEW
EVIEW customers to develop, deploy, and manage their own
Cloud computing security challenges and issues discussed by applications, without insta
installing
lling any of these platforms or
various researchers. The Cloud Computi
Computinngg Use Cases group locaall machines. The PaaS model may
support tools on their loc
[4] discusses the different use case sc enarios and related
scenari be hosted on top of Iaa IaaS
S model or on top of the cloud
requirements that may exist in the clou cloudd computing model. infrastructures directly. Google Apps and Microsoft
They consider use cases from different ppeerspectives
rspectives including Windows Azure are the m st known PaaS.
moost
customers, developers and security eng ineers. ENISA [5]
engineers - Software-as-a-service (Saa
(SaaS): where cloud providers deliver
investigated the different security risks related to adopting applications hosted on ththee cloud infrastructure as internet-
cloud computing along with the affecte affectedd assets, the risks based service for end user
userss, without requiring installing the
likelihood, impacts, and vulnerabilities in cloud computing applications on the custo
custom ers’ computers. This model may
mers’
that may lead to such risks. Similar effort
effortss discussed in “Top be hosted on top of PaaS
PaaS,, IaaS or directly hosted on cloud
Threats to Cloud Computing” by CSA [6 [6]]. Balachandra et al infrastructure. SalesForce CRM is an example of the SaaS
[7] discuss the security SLA’s specificat ions and objectives
specifications provider.
related to data locations, segregation and data recovery.
Kresimir et al [8] discuss high level secu rity concerns in the
security
cloud computing model such as data inte integrity,
grity, payment, and
privacy of sensitive information. Kresimir
Kresimir discussed different
security management standards such as IIT TILIL, ISO/IEC 27001
and Open Virtualization Format (OVF). Me Meiko et al [9] discuss
the technical security issues arising frofromm adopting the cloud
computing model such as XML-attacks, attacks, Browsers’ related
Figure 1: cloud se
service delivery models
attacks, and flooding attacks. Bernd et al [10] discuss the
security vulnerabilities existing in the cloud platform. The Each service delivery model has different possible
authors grouped the possible vulnerabilit
vulnerabilitiies into technology- implementations, as in fi figure 1, which complicates the
gure
related, cloud characteristics -related, secu ity controls- related.
securrity development of standard security model for each service
Subashini et al [11] discuss the security cha llenges of the cloud
challenges delivery model. Moreover, tthhese service delivery models may
service delivery model, focusing on the S SaaaS model. CSA [6] coexist in one cloud platfor
platformm leading to further complication of
discusses critical areas of cloud computin
computingg. They deliver a set the security management pro
proccess.
of best practices for the cloud provider, coconnsumers
sumers and security IV. CLOUD COMPUTING CHARACTERSTICS AND
OMPUTING
vendors to follow in each domain. CSA CSA published a set of SECURITY IMPLICATIONS
SECURITY
detailed reports discussing for some of the
thessee domains.
In our research we did a deep investigation
investigation in the cloud model utilization of resources, cloud providers
To achieve efficient utilizati
to identify the root causes and key particip
participating
ating dimensions in need to increase their reso urce utilization while decreasing
resource
such security issues/problems discussed by by the previous work. cost. At the same time cons
consu mers need to use resources as far
umers
This will help better to understand the problem and deliver as needed while being able to increase or decrease resources
solutions. consumption based on actu actuaall demands. The cloud computing
model meets such needs via via a win-win solution by delivering
III. THE CLOUD COMPUTING AR
ARC HITECTURE AND
CHI two key characteristics: mu multi-tenancy and elasticity. Both
SECURITY IMPLICATIO
IMPLICATIONNS characteristics turn out to have serious implications on the
The Cloud Computing model has three servservice
ice delivery models cloud model security.
and main three deployment models [1 [1]]. The deployment Multi-tenancy implies sha sharring
ing of computational resources,
models are: (1) Private cloud: a cloud pplatform
latform is dedicated storage, services, and appli ations with other tenants. Multi-
appliccations
for specific organization, (2) Public clou
cloud d: a cloud platform tenancy has different realiza
realizattion
ion approaches as shown in figure
available to public users to register an andd use the available 2. In approach 1, each tenan
tenantt has their own dedicated instance
infrastructure, and (3) Hybrid cloud: a prprivate
ivate cloud that can with their own customizati
customizations (customization may include
extend to use resources in public clouds. Public clouds are the special development to meemeett customer needs). In approach 2,
becausee they are available
most vulnerable deployment model becaus each tenant uses a dedicated
dedicated instance, like approach 1, while
for public users to host their services who may be malicious all instances are the same but with different configurations
users. (adjustment of application parameters or interfaces). In
The cloud service delivery models, as in fig
figure1,
ure1, include: approach 3, all tenants shar
sharee the same instance with runtime
- Infrastructure-as-a-service (IaaS): whwhereere cloud providers configuration (the applicati
applicatioonn is divided into core application
deliver computation resources, storage and network as an component and extra compon
components that are loaded based on the
internet-based services. This service mmoodel
del is based on the current tenant requests – similar to SalesForce.com). In
virtualization technology. Amazon EC2 is the most familiar approach 4 tenants are directed to a load balancer that redirects
IaaS provider. tenants requests to a suitable instance based on current
- Platform-as-a-service (PaaS): where clocloudud providers deliver instances load. Approaches 3 and 4 are the most risky as
platforms, tools and other business services that enable tenants are coexisting on th thee same process in memory and
hardware. This sharing of resources violate
violatess the confidentiality covers the platform layers (such as application servers, web
ed for secure multi-
of tenants’ IT assets which leads to the nneeed tools), and APIs and Services layers.
servers, IDEs, and other tool
tenancy. To deliver secure multi-tenanc
tenancy there should be The PaaS layer depends on the virtualization of resources as
isolation among tenants’ data (at res rest, processing and delivered by IaaS. The SaaSaaS S model covers applications and
transition) and location transparency where tenants have no for end users, as shown in figure 3.
services offered as a service for
knowledge or control over the specific
specific location of their The SaaS layer depends oon n a layer of platforms to host the
resources (may have high level control on data location such as services and a layer of vi virrtualization
tualization to optimize resources
country or region level), to avoid planned attacks that attempt rvices to multi-tenant.
utilization when delivering sseervices
to co-locate with the victim assets [12]. In IaaS, isolation
should consider VMs’ storage, processi
processinng, memory, cache
memories, and networks. In PaaS, isol isolation
ation should cover
isolatation among running services and APIs’ calls. In SaaS,
isolation should isolate among transactio
transactionnss carried out on the
same instance by different tenants and tenan
tenants’ data.

Figure 3: Cloud co
computing model layers
This deep dependency stac
stackk of cloud objects complicates the
cloud security problem as the security of each object/layer
depends on the security of the lower objects/layers.
Furthermore, any breach to any cloud objects will impact the
Figure 2: Multi-tenancy approa
approaches [13]
security of the whole cloud platform. Each cloud layer/object
Elasticity implies being able to scale up or down resources has a set of security requirements and vulnerabilities so it
assigned to services based on the current demand. Scaling up requires a set of security cco ntrols to deliver secured service.
ontrols
and down of tenant’s resources gives the opportunity to other This results in a huge numb
numbeerr of security controls that needs to
tenants to use the tenant previously assig
assignedned resources. This be managed. Moreover, man aging such heterogeneous security
anaging
may lead to confidentiality issues. For example, tenant A controls to meet security ne
neeeds
ds is a complex task, taking into
scaled down so it releases resources, these
these resources are now account conflicts among the security requirements and among
assigned to tenant B who in turn use it to deduce the previous security controls at each layer. This may result in an
problem between DNS and
contents of tenant A (similar to lag problem inconsistent security model. Hence, a unified security control
includess a service placement
DNS cache). Moreover, Elasticity include management module is required. This module should
engine that maintains a list of the availabl
availablee resources from the coordinate and integrate am ong the various layers’ security
among
provider’s offered resources pool. This li lisstt is used to allocate controls based on security nee
needs.
resources to services. Such placemenplacementt engines should
incorporate cloud consumers’ security and and legal requirements VI. CLOUD COMPUTING
OMPUTING STAKEHOLDERS AND SECURITY
such as avoid placing competitors service
servicess on the same server, IMPLICATIONS
IMPLICATIONS
data location should be within the tenants’ country boundaries. The cloud computing model has different involved
Placement engines may include a migration strategy where stakeholders: cloud prov provider
ider (an entity that delivers
services are migrated from physical host to another or from infrastructures to the cloud consumers), service provider (an
cloud to another in order to meet dem demandsands and efficient entity that uses the cl cloud infrastructure to deliver
utilization of the resources. This migration strategy should take applications/services to end users), and service consumer (an
into account the same security constr aints. Furthermore,
constraints. entity that uses services hosted on the cloud infrastructure).
security requirements defined by service cconsumers
onsumers should be Each stakeholder has the their own security management
migrated with the service and initiates a process to enforce systems/processes and each one has their own expectations
security requirements on the new environ ment, as defined by
environment (requirements) and capabilities (delivered) from/to other
cloud consumers, and updates the current cl cloud security model. stakeholders. This leads to: (1) A set of security requirements
defined on a service by diffe rent tenants that may conflict with
different
V.CLOUD COMPUTING’S DEEP DEPEND ENNCIES STACK
DEPENDENNCIES each other. So security config urations of each service should be
configurations
The cloud computing model depends oonn a deep stack of maintained and enforced on the service instances level and at
dependent layers of objects (VMs, APIs, Services and runtime taking into accou
account nt the possibility of changing
Applications) where the functionality and security of a higher requirements based on curre nt consumers’ needs to mitigate
current
layer depends on the lower ones. The IaaS
IaaS model covers cloud new risks; (2) Providers and
and consumers need to negotiate and
physical infrastructure layer (storage, nenettworks
works and servers), securit y properties. However, no standard
agree on the applied security
virtualization layer (hypervisors), and virtualized
virtualized resources security specification notatio
notationsns are available that can be used
layer (VMs, virtual storage, virtual network s). The PaaS model
networks). by the cloud stakeholders ttoo represent and reason about their
offered/required security properties; and (3) Each stakeholder vulnerabilities, or even the vSwitch software which result in
has their own security management processes used to define network-based VM attacks.
their assets, expected risks and their impacts, and how to Securing VM boundaries - VMs have virtual boundaries
mitigate such risks. Adopting cloud model results in losing compared with to physical server ones. VMs that co-exist on
control from both involved parties, including cloud providers the same physical server share the same CPU, Memory, I/O,
(who are not aware of the contents and security requirements of NIC, and others (i.e. there is no physical isolation among VM
services hosted on their infrastructures) and cloud consumers resources). Securing VM boundaries is the responsibility of
(who are not able to control neither on their assets security nor the cloud provider.
on other services sharing the same resources). Security SLA Hypervisor security - a hypervisor is the “virtualizer” that maps
management frameworks represent part of the solution related from physical resources to virtualized resources and vice versa.
to security properties specification, enforcement and It is the main controller of any access to the physical server
monitoring. However, SLAs still don’t cover security attributes resources by VMs. Any compromise of the hypervisor violates
in their specifications [14]. Moreover, SLAs are high level the security of the VMs because all VMs operations become
contracts where the details of the security policies and security traced unencrypted. Hypervisor security is the responsibility of
control and how to change at runtime are not included. cloud providers and the service provider. In this case, the SP is
On the other side, cloud providers are not able to deliver the company that delivers the hypervisor software such as
efficient and effective security controls because they are not VMware or Xen.
aware of the hosted services’ architectures. Furthermore, cloud B. PaaS Security Issues
providers are faced with a lot of changes to security SOA related security issues – the PaaS model is based on the
requirements while having a variety of security controls Service-oriented Architecture (SOA) model. This leads to
deployed that need to be updated. This further complicates the inheriting all security issues that exist in the SOA domain such
cloud providers’ security administrators’ tasks. Transparency as DOS attacks, Man-in-the-middle attacks, XML-related
of what security is enforced, what risks exist, and what attacks, Replay attacks, Dictionary attacks, Injection attacks
breaches occur on the cloud platform and the hosted services and input validation related attacks [9, 16]. Mutual
must exist among cloud providers and consumers. This is what authentication, authorization and WS-Security standards are
is called “trust but verify” [15], where cloud consumers should important to secure the cloud provided services. This security
trust in their providers meanwhile cloud providers should issue is a shared responsibility among cloud providers, service
deliver tools to help consumers to verify and monitor security providers and consumers.
enforcements. API Security - PaaS may offer APIs that deliver management
functions such as business functions, security functions,
VII. CLOUD COMPUTING SERVICE DELIVERY MODELS
application management, etc. Such APIs should be provided
AND SECURITY IMPLICATIONS
with security controls and standards implemented, such as
We summarize the key security issues/vulnerabilities in each OAuth [17], to enforce consistent authentication and
service delivery model. Some of these issues are the authorization on calls to such APIs. Moreover, there is a need
responsibility of cloud providers while others are the for the isolation of APIs in memory. This issue is under the
responsibility of cloud consumers. responsibility of the cloud service provider.
A. IaaS Issues C. SaaS Security Issues
VM security – securing the VM operating systems and In the SaaS model enforcing and maintaining security is a
workloads from common security threats that affect traditional shared responsibility among the cloud providers and service
physical servers, such as malware and viruses, using traditional providers (software vendors). The SaaS model inherits the
or cloud-oriented security solutions. The VM’s security is the security issues discussed in the previous two models as it is
responsibility of cloud consumers. Each cloud consumer can built on top of both of them including data security
use their own security controls based on their needs, expected management [11] (data locality, integrity, segregation, access,
risk level, and their own security management process. confidentiality, backups) and network security.
Securing VM images repository - unlike physical servers VMs Web application vulnerability scanning - web applications to
are still under risk even when they are offline. VM images can be hosted on the cloud infrastructure should be validated and
be compromised by injecting malicious codes in the VM file or scanned for vulnerabilities using web application scanners [18].
even stole the VM file itself. Secured VM images repository is Such scanners should be up to date with the recently
the responsibilities of the cloud providers. Another issue discovered vulnerabilities and attack paths maintained in the
related to VM templates is that such templates may retain the National Vulnerability Database (NVD) and the Common
original owner information which may be used by a new Weaknesses Enumeration (CWE) [19]. Web application
consumer. firewalls should be in place to mitigate existing/discovered
Virtual network security - sharing of network infrastructure vulnerabilities (examining HTTP requests and responses for
among different tenants within the same server (using vSwitch) applications specific vulnerabilities). The ten most critical web
or in the physical networks will increase the possibility to applications vulnerabilities in 2010 listed by OWASP [20] are
exploit vulnerabilities in DNS servers, DHCP, IP protocol injection, cross site scripting (Input validation) weaknesses.
Web application security miss-configuration and breaking - approaches have a major problem related to encryption key
web application security miss-configuration or weaknesses in management i.e. how to securely generate, store, access and
application-specific security controls is an important issue in exchange secrete keys. Moreover, PaaS requires application
SaaS. Security miss-configuration is also very critical with keys for all APIs and service calls from other applications. The
multi-tenancy where each tenant has their own security applications’ keys must be maintained securely along with all
configurations that may conflict with each other leading to other credentials required by the application to be able to
security holes. It is mostly recommended to depend on cloud access such APIs.
provider security controls to enforce and manage security in a C. Security Management
consistent, dynamic and robust way. Based on the huge number of cloud stakeholders, the deep
D. Cloud Management Security Issues dependency stack, and the large number of security controls to
The Cloud Management Layer (CML) is the “microkernel” that deliver security requirements, the cloud security management
can be extended to incorporate and coordinate different becomes a more complicated research problem. Security
components. The CML components include SLA management, management needs to include security requirements and
service monitoring, billing, elasticity, IaaS, PaaS, SaaS services policies specifications, security controls configurations
registry, and security management of the cloud. Such a layer is according to the policies specified, and feedback from the
very critical since any vulnerability or any breach of this layer environment and security controls to the security management
will result in an adversary having control, like an administrator, and the cloud stakeholders. Security management should
over the whole cloud platform. This layer offers a set of APIs function as a plug-in for CML.
and services to be used by client applications to integrate with D. Secure Software Development Lifecycle
the cloud platform. This means that the same security issues of The secure software development lifecycle (SDLC with
the PaaS model apply to the CML layer as well. security engineering activities) includes elicitation of the
E. Cloud Access Methods Security Issues security requirements, threat modeling, augmentation of
Cloud computing is based on exposing resources over the security requirements to the systems models and the generated
internet. These resources can be accessed through (1) web code consequently. The cloud based applications will involve
browsers (HTTP/HTTPS), in case of web applications - SaaS; revolution in the lifecycles and tools used to build secure
(2) SOAP, REST and RPC Protocols, in case of web services systems. The PaaS provides a set of reusable security enabling
and APIs – PaaS and CML APIs; (3) remote connections, VPN components to help developing secured cloud-based
and FTP in case of VMs and storage services – IaaS. Security applications. Also security engineering of the cloud-based
controls should target vulnerabilities related to these protocols application should change to meet new security requirements
to protect data transferred between the cloud platform and the imposed on such systems. Applications should support
consumers. adaptive security (avoiding hardcoded security) to be able to
meet vast range of consumers’ security requirements. Adaptive
VIII. CLOUD COMPUTING SECURITY ENABLERS
application security is based on externalizing/delegating the
A. Identity & Access Management (IAM) and Federation security enforcement and applications security management to
Identity is a core of any security aware system. It allows the the cloud security management, cloud security services and
users, services, servers, clouds, and any other entities to be security controls.
recognized by systems and other parties. Identity consists of a E. Security-Performance tradeoff optimization
set of information associated with a specific entity. This The cloud computing model is based on delivering services
information is relevant based on context. Identity should not using SLAs. SLAs should cover objectives related to
disclose user personal information “privacy”. Cloud platforms performance, reliability, and security. SLAs also define
should deliver or support a robust and consistent Identity penalties that will be applied in case of SLA violation.
management system. This system should cover all cloud Delivering high security level, as one of SLA objectives, means
objects and cloud users with corresponding identity context consuming much more resources that impact on the
information. It should include: Identity Provisioning and de- performance objective (the more adopted security tools and
provisioning, identity information privacy, identity linking, mechanism, the worst the impact on the performance of the
identity mapping, identity federation, identity attributes underlying services). Cloud management should consider the
federation, single sign on, authentication and authorization. trade-off between security and performance using utility
Such system should adopt existing standards, such as SPML, functions for security and performance (least security unless
SAML, OAuth, and XACML, to securely federate identities stated otherwise). Moreover, we should focus on delivering
among interacting entities within different domains and cloud adaptive security where security controls configurations are
platforms. based on the current and expected threat level and considering
B. Key Management other tradeoffs.
Confidentiality is one of key objectives of the cloud computing F. Federation of security among multi-clouds
security (CIA triad). Encryption is the main solution to the When a consumer uses applications that depend on services
confidentiality objective, for data, processes and from different clouds, he will need to maintain his security
communications. Encryption algorithms either symmetric key- requirements enforced on both clouds and in between. The
based or asymmetric are key-based. Both encryption same case when multiple clouds integrate together to deliver a
bigger pool of resources or integrated services, their security loop will measure the security status to help improving the
requirements needs to be federated and enforced on different current cloud security model and keeping cloud consumers
involved cloud platforms. aware with their assets’ security status (applying the trust but
verify concept).
IX. CONCLUSION
REFERENCES
The cloud computing model is one of the promising computing [1] Peter Mell, and Tim Grance, "The NIST Definition of Cloud Computing,"
models for service providers, cloud providers and cloud 2009, http://www.wheresmyserver.co.nz/storage/media/faq-files/cloud­
consumers. But to best utilize the model we need to block the def-v15.pdf, Accessed April 2010.
existing security holes. Based on the details explained above, [2] Frank Gens, Robert P Mahowald and Richard L Villars. (2009, IDC Cloud
Computing 2010.
we can summarize the cloud security problem as follows: [3] IDC, "IDC Ranking of issues of Cloud Computing model," ed, 2009,
• Some of the security problems are inherited from the used http://blogs.idc.com/ie/?p=210, Accessed on July 2010.
technologies such as virtualization and SOA. [4] Cloud Computing Use Case Discussion Group, "Cloud Computing Use
Cases Version 3.0," 2010.
• Multi-tenancy and isolation is a major dimension in the [5] ENISA, "Cloud computing: benefits, risks and recommendations for
cloud security problem that requires a vertical solution from information security," 2009,
the SaaS layer down to physical infrastructure (to develop http://www.enisa.europa.eu/act/rm/files/deliverables/cloud-computing­
physical alike boundaries among tenants instead of virtual risk-assessment, Accessed On July 2010.
[6] Cloud Security Alliance (CSA). (2010). Available:
boundaries currently applied). http://www.cloudsecurityalliance.org/
• Security management is very critical to control and manage [7] Balachandra Reddy Kandukuri, Ramakrishna Paturi and Atanu Rakshit,
this number of requirements and controls. "Cloud Security Issues," in Proceedings of the 2009 IEEE International
• The cloud model should have a holistic security wrapper, as Conference on Services Computing, 2009, pp. 517-520.
[8] Kresimir Popovic , Zeljko Hocenski, "Cloud computing security issues and
shown in figure 3, such that any access to any object of the challenges," in The Third International Conference on Advances in
cloud platform should pass through security components Human-oriented and Personalized Mechanisms, Technologies, and
first. Services, 2010, pp. 344-349.
Based on this discussion we recommend that cloud computing [9] Meiko Jensen, Jörg Schwenk, Nils Gruschka and Luigi Lo Iacono, "On
Technical Security Issues in Cloud Computing," in IEEE ICCC, Bangalore
security solutions should: 2009, pp. 109-116.
• Focus on the problem abstraction, using model-based [10] Bernd Grobauer, Tobias Walloschek and Elmar Stöcker, "Understanding
approaches to capture different security views and link such Cloud-Computing Vulnerabilities," IEEE Security and Privacy, vol. 99,
2010.
views in a holistic cloud security model.
[11] S. Subashini, ,Kavitha, V., "A survey on security issues in service delivery
• Inherent in the cloud architecture. Where delivered models of cloud computing," Journal of Network and Computer
mechanisms (such as elasticity engines) and APIs should Applications, vol. In Press, Corrected Proof.
provide flexible security interfaces. [12] Thomas Ristenpart, Eran Tromer, Hovav Shacham, Stefan Savage, "Hey,
you, get off of my cloud: exploring information leakage in third-party
• Support for: multi-tenancy where each user can see only his compute clouds," presented at the Proceedings of the 16th ACM
security configurations, elasticity, to scale up and down conference on Computer and communications security, Chicago, Illinois,
based on the current context. USA, 2009.
• Support integration and coordination with other security [13] Microsoft. (2006, October, 2010). Multi-Tenant Data Architecture.
Available: http://msdn.microsoft.com/en-us/library/aa479086.aspx
controls at different layers to deliver integrated security. [14] Amazon. October, 2010). Amazon EC2 SLA. Available:
• Be adaptive to meet continuous environment changes and http://aws.amazon.com/ec2-sla/
stakeholders needs. [15] D. K. Holstein, , Stouffer, K., "Trust but Verify Critical Infrastructure
Cyber Security Solutions," in HICSS 2010, pp. 1-8.
X. FUTURE WORK [16] Z. Wenjun, "Integrated Security Framework for Secure Web Services," in
IITSI 2010, pp. 178-183.
We are investigating in the cloud security management [17] B. Wang, Huang He, Yuan, Liu Xiao, Xi, Xu Jing, Min, "Open Identity
Management Framework for SaaS Ecosystem," in ICEBE '09. pp. 512­
problem. Our objective is to block the hole arise in the security 517.
management processes of the cloud consumers and the cloud [18] F. Elizabeth, , Vadim, Okun, "Web Application Scanners: Definitions and
providers from adopting the cloud model. To be able to resolve Functions," in HICSS 2007, pp. 280b-280b.
such problem we need to: (1) Capture different stakeholders [19] NIST. October, (2010). National Vulnerability Database (NVD).
Available: http://nvd.nist.gov/home.cfm
security requirements from different perspectives and different [20] OWASP. (2010, The Ten Most Critical Web Application Security
levels of details; (2) Map security requirements to the cloud Vulnerabilities. Available:
architecture, security patterns and security enforcement http://www.owasp.org/index.php/OWASP_Top_Ten_Project
mechanisms; and (3) Deliver feedback about the current
security status to the cloud providers and consumers. We
propose to adopt an adaptive model-based approach in tackling
the cloud security management problem. Models will help in
the problem abstraction and the capturing of security
requirements of different stakeholders at different levels of
details. Adaptive-ness will help in delivering an integrated,
dynamic and enforceable cloud security model. The feedback

View publication stats