![](bg8.png)
knowledge there are no further works related with
the quantitative security assessment of CSPs and in
particular aimed to empirically validate their security
metrics with real CSP data. Nevertheless for the sake
of completeness, the rest of this section cites the ef-
forts from other Information Technology fields (e.g,
Web Services and Grid computing) aimed to adopt
and assess SecLAs.
In (Frankova, G. and Yautsiukhin, A., 2007) and
(Krautsevich, L., et.al., 2011), the authors propose a
technique to aggregate security metrics from a web
services’ SecLA, however contrary to our research
they did not propose the techniques to assess their
SecLAs neither empirically validate the proposed
metrics.
The Reference Evaluation Methodology — REM
— (as explained in Section 3.1) was originally pro-
posed in (Casola V., et.al., 2006). The authors intro-
duced a methodology that can be adopted whenever
there is the need of evaluating and comparing secu-
rity SLAs (despite not specifically Cloud-related) ex-
pressed through the use of standard policy languages.
A similar approach was used in (Casola V., et.al.,
2005) and (Casola, V. et.al., 2007) to quantify the se-
curity of a Public Key Infrastructure, based on its Cer-
tificate Policy. The security assessment presented in
our paper has been built above the methodology from
(Casola V., et.al., 2006) by quantitatively evaluating a
Cloud SecLA, but contrary to existing works we have
also contributed with: (i) an additional set of security
metrics to quantitatively assess the CSP, (ii) an ini-
tial Cloud SecLA specification and, (iii) the building
blocks of an architecture aimed to empirically demon-
strate the assessment of the CSP.
In (De Chaves, S. A., et.al., 2010) the authors
highlight the importance of incorporating security
metrics in SLAs and, in particular, of controlling
and monitoring whether the security metrics are met.
However, contrary to our paper, no further details are
provided about the techniques used to represent and
assess these SLAs.
A metric-based approach for assessing the secu-
rity level of Critical Infrastructures was presented
in (Ghani, H., et.al., 2010). In that article the au-
thors define appropriate security metrics to monitor
whether the established security requirements are ful-
filled. Such metrics are also used for the definition
of SLAs that should capture the defined requirements
as well as the guarantees that the system provides to-
gether with the penalties that have to be applied when
such guarantees are not met. The present paper took
into account our previous experiences from (Luna, J.,
et.al., 2011), in order to contribute with the metrics
and testbed to quantitatively assess Cloud SecLAs.
The works presented in (Irvine, C. and Levin, T.,
2001) and (Neto, A., et.al., 2011) are also related
with the assessment metrics presented in this paper.
In (Irvine, C. and Levin, T., 2001), the authors sup-
port the notion of “Quality of Security” (similar to our
SecLA quantification) and the usefulness of security
ranges for quantifying security. Both arguments are
directly related with the security quantification tech-
nique used by our framework (cf. Section 3). In
(Neto, A., et.al., 2011) the authors present one of the
few works related with security benchmarking. From
their perspective trust can be used to benchmark secu-
rity by performing the enumeration and accumulation
of the evidence in favor of security that exists in a sys-
tem, in the form of security mechanisms, processes,
configurations, procedures and behaviors that prevent
the accomplishment of specific threats that the sys-
tem may be subjected to. These notions are directly
related with the use of metrics for benchmarking se-
curity in our CSP assessment.
Finally, from an industrial perspective SecLAs
must also be considered in security management regu-
lations and standards as mentioned in (Monahan, B.
and Yearworth, M., 2008). The Information Tech-
nology Infrastructure Library (ITIL) is one of the
most prestigious good practices regarding IT Services
Management and, some works have been done on
modeling SLAs considering risks that threaten busi-
ness processes and the requisites detailed in the Ser-
vice Level Management module from ITIL (Feglar,
T., 2004). However, contrary to the approach pre-
sented in our paper the techniques for quantitatively
assessing these SecLAs in ITIL were not proposed in
(Monahan, B. and Yearworth, M., 2008) nor (Feglar,
T., 2004).
6 CONCLUSIONS
In this paper we have presented a practical CSP’s
security assessment use case based on (i) the con-
cept of Security Level Agreements and, (ii) a novel
set of quantitative security assessment metrics. This
practical security assessment has been applied to the
CSP’s information stored in the STAR repository of
the Cloud Security Alliance. To the best of our knowl-
edge there are no previous works related with the
quantitative security assessment of CSPs, and in par-
ticular aimed to empirically validate the applicability
of their security metrics with real CSP data.
The obtained results show that the proposed se-
curity assessment can be performed at different lev-
els of the CSP, ranging from the Cloud SecLA-level
to the individual security provision-level. Our results
QuantitativeAssessmentofCloudSecurityLevelAgreements-ACaseStudy
71