A Quantitative Methodology for Security Risk Assessment of
Enterprise Business Processes
Jaya Bhattacharjee, Anirban Sengupta and Chandan Mazumdar
Centre for Distributed Computing, Jadavpur University, Jadavpur, Kolkata, India
Keywords: Business Process, Risk Assessment, Task, Threat, Vulnerability.
Abstract: Business processes help to realize the business objectives of an enterprise. Security breach of business
processes may lead to un-fulfillment of objectives, loss of revenue, and possible shutdown of the
corresponding business venture. Hence, it is important to ensure that the security properties of critical
business processes are protected from attacks and failures. Effective protection mechanisms can be designed
only after identifying security risks to business processes. However, existing methodologies mostly focus on
the detection of risks to individual hardware, software, network and information assets. They do not cater to
risks that are specific to business processes. This paper attempts to address this gap in research by
describing a technique for identifying the components of a business process and quantitatively assessing
their security risks.
1 INTRODUCTION
An enterprise is usually defined as an organization
that has been created to execute one or more
business ventures (ed. Soanes and Stevenson, 2011).
Such ventures are realized with the help of certain
types of activities that are referred to as business
processes (ISO/IEC, 2011). Business processes are
responsible for building / generating products of an
enterprise, as well as for delivering services to
clients. Examples include processes for building a
car, developing software, or providing online money
transfer services to a bank’s customers. Thus, the
very existence of an enterprise depends on the
proper functioning of its business processes.
Security breach of business processes would lead to
un-fulfilment of an enterprise’s security objectives,
loss of revenue, and possibly shutdown of the
corresponding business venture.
Hence, it is extremely important to ensure that
the security properties of critical business processes
are protected from malicious attacks and accidental
failures. However, effective protection mechanisms
cannot be designed without identifying the specific
security risks to business processes. Historically,
there has been lots of research into methodologies
for identifying risks to hardware, software, network
and information assets of an enterprise (Peltier,
2010; Bhattacharjee et al., 2013). These
methodologies are successful in evaluating the risks
to assets that are required for executing business
processes.
However, there is another class of security risks
that exist owing to vulnerabilities within the basic
structure of business processes, and are independent
of the underlying assets of an enterprise. This
includes both internal as well as external risks. For
example, lack of training on use of privileged access
rights (internal risk) could lead to users divulging
their credentials, thus leading to serious security
breaches. Again, high attrition of personnel owing to
better opportunities elsewhere (external risk) could
seriously jeopardize enterprise operations. Barring a
few attempts (ENISA, 2015; MEHARI, 2010), there
has been hardly any significant research for
formulating techniques for the identification of such
business process-specific risks. This leads to
incomplete design of protection mechanisms, thus
exposing an enterprise to various security breaches.
This paper attempts to fill this research gap by
proposing a quantitative methodology for the
assessment of security risks (both internal and
external) to enterprise business processes.
Rest of this paper is organized as follows.
Section 2 presents a survey of related work. Section
3 details the components of a business process and
their inherent relationships. Section 4 describes the
388
Bhattacharjee, J., Sengupta, A. and Mazumdar, C.
A Quantitative Methodology for Security Risk Assessment of Enterprise Business Processes.
DOI: 10.5220/0005739703880399
In Proceedings of the 2nd International Conference on Information Systems Security and Privacy (ICISSP 2016), pages 388-399
ISBN: 978-989-758-167-0
Copyright
c
2016 by SCITEPRESS Science and Technology Publications, Lda. All rights reserved
vulnerabilities in, and threats to, business processes,
and proposes a quantitative methodology for
security risk assessment. While Section 5 discusses
the benefits of the proposed methodology, an
illustrative case study is included in Section 6.
Finally, Section 7 concludes the paper. It may be
noted that some of the terms and definitions used in
Section 4 have been adopted from Bhattacharjee et
al., (2013); Bhattacharjee et al., (2014).
2 RELATED WORK
Several researchers have discussed the security
issues of enterprise business processes. Some of
them have also proposed techniques for modelling
risk-aware business processes. We discuss some of
the significant contributions in these areas.
Marchesini and Viganò (2011) discussed an
approach for the formal analyses of business
processes that need to comply with security
requirements like authorization constraints, or
separation or binding of duties. They observed that a
business process has two levels: the workflow level
dealing with the control of the flow (and the
manipulation of data) and the policy management
level describing access rules and permissions. They
introduced a notion of knowledge hierarchy within
the entities of a business process that is involved in
the interaction among workflow and policy
management levels. An entity’s state of knowledge
represents the entity’s view of the business process.
The authors have attempted to include information
about sets of security-critical tasks at different levels
of hierarchy that can be used to control the process
execution and enforce security properties.
Armando and Ponta (2011) discussed about
authorization requirements of security-sensitive
business processes. In a business process, agents can
be dynamically delegated to perform tasks they were
not initially authorized to execute. Considering this,
they proposed a new approach for the specification
and automatic analysis of security-sensitive business
processes. They have used model checking to
analyze the specification of the workflow and of the
associated security policies separately.
Both of the above methods address the
authorization aspects of business processes. They
have not considered other important security issues
like confidentiality, integrity and availability
requirements of processes.
Lowis and Accorsi (2011) proposed a method to
search and analyze the vulnerabilities of SOA-based
business processes and services. They have proposed
six attack effects for business processes
corresponding to confidentiality, integrity and
availability parameters: start, stop, steer, split, spot
and study. An attacker can start or stop a process and
may harm availability. He can steer or split a process
and can harm integrity. Finally, ability of an attacker
to spot or study a process can harm its
confidentiality. Though the method analyzes
vulnerabilities within business processes, it does not
explicitly address threats or compute risk values.
Tjoa et al., (2011) proposed a formal model that
considers relations between threats, detection
mechanisms, safeguards, recovery measures and
their effects on business processes. Business process
is represented by a set of resources, activities and
their attributes. Then threats to the attributes of
different elements of business process are identified
and their preventive, blocking and reactive measures
are stated by the model.
Khanmohammadi and Houmb (2010) proposed a
business process based risk assessment methodology
and focused on business goals rather than assets.
Business and their control processes are identified
during the initial phase. Then vulnerabilities within
these processes are identified and threats to those
vulnerabilities are analyzed. Finally, risk is
computed considering the degree of exposure of
vulnerabilities, effects of installed security controls,
threat levels and process value.
Jakoubi et al., (2010) presented a technique for
risk-aware business process management. It consists
of five distinct phases: Perform Program
Management, Determine As-Is Situation, Re-
engineer Processes, Implement Processes and
Review and Evaluate. However, the methodology is
mostly verbose and does not suggest any
quantitative or formal technique for the computation
of risks to business processes.
The above discussion shows that though some
techniques for analyses of risks to business
processes have been presented, most of them are
verbose, qualitative approaches. They do not strive
to model such risks quantitatively. Besides, there is
also a lack of understanding of the internal structure
of a business process that is so essential for
developing a quantitative approach. Though BPMN
(Business Process Model and Notation) provides
graphical notations for enabling enterprises to model
their business processes (OMG, 2011), there has
been limited adoption of the approach in case of
information security. The research presented in this
paper attempts to fill these gaps by describing a
technique for identifying the components of a
business process and computing values of their
security risks.
A Quantitative Methodology for Security Risk Assessment of Enterprise Business Processes
389
3 BUSINESS PROCESS
Business process, as stated above, refers to the
primary activities that are essential for achieving the
business objectives of an enterprise. It consists of a
set of tasks (or activities) that can be arranged in a
linear sequence, in parallel, as a conditional
structure, or repetitive (loop) structure. These are
depicted in Figures 1(a), 1(b), 1(c) and 1(d),
respectively, where T
i
denote tasks and C
j
denote
specific conditions. In addition to depicting simple
if…else kind of instances, conditional task structures
may also be used to represent situations where
alternate paths exist. For example, a bank may offer
cash withdrawal facilities via multiple means:
through cheques, withdrawal forms, or ATM. The
underlying procedure for each of them may be
similar, though they differ in their actual
implementations. Such a situation illustrates
alternate task structures, where the actual path
traversed would depend on the customer. On the
other hand, parallel task structures are used in cases
where multiple tasks can be executed
simultaneously. This can occur when such tasks (or
their predecessors) have the same ancestor, and they
(or their successors) have the same descendant. In
Figure 1(b), T
9
and T
10
are parallel tasks; both of
them have task T
8
as their ancestor, and task T
11
as
their descendant.
To illustrate the concept of business processes
and tasks lucidly, let us consider the loan
management process of a bank (Figure 2). This
process contains different types of task structures
(sequential, parallel, conditional and repetitive) as
stated above. A task can be as simple as writing a
standard, pre-defined report or it may consist of
complex functions as illustrated by the “loan
processing” task in Figure 2. Thus, a task can be
viewed as comprising of a set of functions or sub-
routines, all of which must be carried out to
complete the corresponding task. The structure of an
individual task is shown in Figure 3. In order to
execute the functions of a task, three types of
conditions may need to be fulfilled:
Figure 1(a): Sequential
tasks.
Figure 1(b): Parallel tasks.
Figure 1(c): Conditional
tasks.
Figure 1(d): Loop.
Figure 2: Loan management process.
Figure 3: Structure of a task.
1) Completion of Pre-requisites. Some tasks may
need to be triggered by certain other tasks /
business processes. Similarly, in a sequential task
ICISSP 2016 - 2nd International Conference on Information Systems Security and Privacy
390
structure, a task, say t
i
, can commence only after
task t
i-1
has been completed. These instances
illustrate the fact that there might be some pre-
requisites (usually other tasks, as shown here)
that need to be fulfilled before a specific task can
be executed.
2) Availability of Resources. One or more
resource(s) may be required to carry out a task.
Resources can be of several types: infrastructure
(hardware, software, physical space or site,
network components, etc.), money, time, agent
(human or software agent who will execute the
task), and competence. Competence can be
viewed as a function of knowledge, skills, and
attitude. An agent must possess the requisite
competence to be able to carry out the task with
the help of available infrastructure. Money may
be needed to acquire the necessary infrastructure
or employ agents. Certain amount of time is
usually required to carry out a task. This may
include the time needed to prepare the
infrastructure and acquire relevant competence,
besides the time spent in actual execution of the
task.
3) Availability of Input. A task may need one or
more input(s) to execute its functions. Such input
may either be received from the environment, or
produced as output by another task that has
already been completed. Generally, inputs can be
of two types: data items and control items. Since,
control items have already been taken care of by
pre-requisites (triggers or natural control flow
owing to completion of the previous task), only
data items are considered as input elements in the
proposed task structure.
It may be noted that it is not necessary that all of
the three types of conditions described above are
required by each and every task. A particular
task might not need some (or any) of these
conditions. The outcome of a task might be
tangible or intangible. Different types of
outcome are as follows:
1) Data Item (Output). A task might produce one
or more data items (report, list, etc.) as output.
These data items may either serve as input to
another task, or may be used by an enterprise for
other purposes. For example, there may be a task
to generate the balance-sheet of an enterprise that
is subsequently published for public viewing.
2) Trigger. As stated above, the completion of a
task might trigger the commencement of another.
Thus, an outcome of a task may simply be a
signal for another task / process to begin
execution.
3) Acknowledgement. Another type of outcome
may be an acknowledgement that the task has
completed its execution. This may be transmitted
to the process / task that had initiated the task in
question, based on which further actions may be
performed.
Thus, task t
i
can be represented as
t
i
= {inp, pr, res, fn, out}
(1)
where, inp denotes a set of inputs, pr denotes pre-
requisite set, res denotes resource set, fn denotes a
set of functions and out denotes the outcome set.
This implies that, in essence, task comprises of a set
of functions that transform inputs to corresponding
outcomes, with the help of resources and pre-
requisites.
Business process, BP
l
, can be represented as
BP
l
= {t
i
,  i > 0}
(2)
where, t
i
denotes a set of tasks (as defined above)
and denotes a flow relation (Sun et al., 2006;
Denning, 1976) between the tasks. The flow relation
signifies the order of execution of tasks within a
business process, and can be represented as t
i
t
j
,
which means that task t
j
should follow task t
i
. As
stated above, the flow relation between two tasks
can be either one of the following: sequential,
iterative, conditional or parallel. It is absolutely
essential for the tasks to follow the pre-defined flow
relation, else the outcomes of the tasks and hence, of
the business process, may either be incorrectly
generated, or not generated at all. For example,
Figure 2 shows the following flow relation between
tasks: “Verification of applicant” “Loan
processing” “Loan sanction” “Loan
disbursement”. Now, if it so happens that the flow
relation is erroneously altered and the following
sequence of tasks is executed: “Loan processing”
“Loan sanction” “Loan disbursement”
“Verification of applicant”. This could lead to
disastrous consequences where the loan might get
disbursed to an ineligible, even fraudulent, applicant.
Business processes are usually perceived as the
most valuable primary assets of an enterprise
(ISO/IEC, 2011). If a business process gets
adversely affected, the very existence of the
enterprise might be at stake. Hence, it is important to
derive a methodology for computing values of the
components of business processes and identifying
probable risks, so that appropriate security measures
can be deployed to protect them.
The technique for deriving values of tasks and
assessing risks to business processes is described in
the following section.
A Quantitative Methodology for Security Risk Assessment of Enterprise Business Processes
391
4 RISKS TO BUSINESS PROCESS
The risk assessment methodology begins by
computing security values of the component tasks of
business processes.
4.1 Valuation of Tasks
The proposed valuation technique hinges on
identifying the confidentiality, integrity and
availability requirements of a task. Confidentiality
(C) of an entity refers to the need for protecting its
secrecy. In case of a task, confidentiality may mean
that unauthorized entities should not gain access to
the elements of a task. It could also mean that
unauthorized entities should not even be aware of
the existence of a task. In technical terms, the former
condition is referred to as “the ability to study” the
task, while the latter is known as “the ability to spot”
the task (Lowis and Accorsi, 2011). Enterprises like
defense and space research organizations, which
have very strict requirements for task confidentiality,
need to implement controls wherein unauthorized
users are not able to detect even the existence of the
task. It may be possible for an entity to “study” the
following elements of a task:
Input
Pre-requisites
Resources
Functions
Outcome
Table 1: Assigning confidentiality values of tasks.
C-value Interpretation
5 Only authorized entities can spot the task
4
Unauthorized entities can spot the task, but may
not study any of its elements (input, pre-
requisites, resources, functions, outcome)
3
Unauthorized entities may study the outcome of a
task
2
Unauthorized entities may study all task elements
except its functions
1 Unauthorized entities may study all task elements
The above elements may be used to define a
graded system for assigning confidentiality values of
tasks. Table 1 shows the proposed C-value
assignment scheme. The following assumptions
have been made:
If an entity is not allowed to “spot” a task, it
implies that the entity will also not be able to
“study” it;
“Study” of the outcome of a task is considered
less critical than “study” of any other element of
the task; and
“Study” of the functions of a task is considered
to be the most critical activity.
Integrity (I) requirement of a task refers to the need
for protecting its accuracy and completeness. A task
should exist in its entirety and all of its component
elements should be accurate. Specifically, this
implies that all inputs, pre-requisites, resources and
functions of the task are present and are fulfilling
their objectives correctly; similarly, all outcomes of
the task are being generated correctly as per
requirements. Table 2 shows the proposed I-value
assignment scheme.
Table 2: Assigning integrity values of tasks.
I-value Interpretation
5
All elements of the task (input, pre-requisites,
resources, functions, outcome) are always
accurate and complete
4
All elements of the task remain accurate and
complete whenever needed for execution
3
Whenever the task is executed, its outcome is
obtained accurately (but may not be complete)
2
The outcome of the task is obtained accurately
majority of the time (during a period of
measurement)
1 Integrity requirement of the task is negligible
Availability (A) of a task means that the task
should remain accessible and usable. The former
implies that authorized entities are able to access the
task without hindrance (e.g. there is no “denial of
service”), while the latter means that the task can be
used to fulfill its intended objective. Table 3 shows
the proposed A-value assignment scheme.
It is important to note that value adjustment due
to inter-task dependencies is not carried out
explicitly in the proposed technique. This is owing
to the fact that such adjustment has already been
taken care of implicitly during computation of
values of individual tasks. For example, C-value
considers all elements of a task that includes its pre-
requisites and inputs. A task, say t
j
, can be
dependent on another task, say t
i
, in one or more of
the following ways:
the outcome of t
i
may serve as input to t
j
;
the completion of t
i
may serve as a pre-requisite
for the initiation of t
j
.
As is obvious, both of the above have been
considered during the valuation of task t
j
.
After computation of the security values of tasks,
vulnerabilities within the tasks, and their
corresponding threats, need to be identified.
ICISSP 2016 - 2nd International Conference on Information Systems Security and Privacy
392
Table 3: Assigning availability values of tasks.
A-value Interpretation
5
All required elements of the task (input, pre-
requisites, resources, functions, or outcome) are
always accessible and usable
4
All required elements of the task are always
accessible, and remain usable whenever needed
for execution
3
All required elements of the task remain
accessible and usable whenever needed for
execution
2
All required elements of the task remain
accessible and usable majority of the time (during
a period of measurement)
1 Availability requirement of the task is negligible
4.2 Vulnerabilities
Vulnerability is defined as an inherent weakness in
an entity that can be exploited by threat(s) to breach
security of that entity (ISO/IEC, 2014). Improper
configuration of tasks and lack of checks and
controls may expose the different task elements to
security risks. Following are some examples of
vulnerabilities within tasks:
Non-availability of manpower may hamper task
execution;
Lack of training of concerned personnel may
lead to incorrect execution of tasks;
Improperly configured infrastructure may hinder
task execution, or lead to unauthorized access
and attacks;
Lack of input validation may result in improper
task execution, or generation of erroneous
outcome;
Lack of monitoring and management of pre-
requisites may lead to non-initiation of a task;
Improperly configured functions may generate
erroneous outcome, or may not produce any
outcome;
Lack of output validation may cause leakage of
confidential information to unauthorized entities.
Based on the component elements of a task,
vulnerabilities may be categorized as follows:
Input vulnerabilities;
Vulnerabilities owing to pre-requisites;
Resource vulnerabilities;
Function vulnerabilities; and
Output vulnerabilities.
It is important to devise means for identifying
vulnerabilities in tasks and analyzing their
criticalities. All identified vulnerabilities may not be
equally critical. The gravity of vulnerabilities may
be recognized from the values of two vital attributes,
namely severity and exploitability. These are
described in the following sub-sections.
4.2.1 Severity
Severity (Sev) of vulnerability indicates how bad the
vulnerability is (Bhattacharjee et al., 2013). It is
determined by the amount and type of impact that
can occur if the vulnerability is successfully
exploited by corresponding threat(s). Types of
impact may be broadly classified as loss of
confidentiality, integrity and/or availability of tasks.
Breach of these security parameters may be caused
by unauthorized entities when they gain illegal
access as follows:
Confidentiality of a task may be breached if
unauthorized entities can “spot” or “study” the
elements of the task;
Integrity may be breached if entities can illegally
“alter” or “obliterate” the elements of the task;
Availability may be breached if entities can
illegally “alter” the task elements or make them
“inaccessible”.
It is important to note that vulnerabilities within the
elements of a task might also jeopardize the security
of other, dependent tasks. For example, consider the
case where the outcome of a task, say t
i
, is later used
as an input to another task, say t
j
. A vulnerability in
t
i
is exploited by a threat leading to illegal alteration
of its outcome, and hence causing breach of its
integrity. Moreover, this also impacts the input to t
j
,
which may render it useless, leading to loss of
availability of t
j
.
In the proposed methodology, severity is
computed on a 5-point scale, as shown in Table 4.
The assumption stated in Section 4 holds here as
well; that is, if an entity can “study” a task, it implies
that the entity will also be able to “spot” it.
Table 4: Severity values of vulnerabilities.
Severity Interpretation
Very
High (5)
Vulnerability in task allows unauthorized
entities to study and/or alter/obliterate
element(s) of multiple tasks and/or make them
inaccessible.
High (4)
Vulnerability in task allows unauthorized
entities to study and/or alter/obliterate
element(s) of the task and/or make them
inaccessible.
Medium
(3)
Vulnerability in task allows unauthorized
entities to spot element(s) of multiple tasks.
Low (2)
Vulnerability in task allows unauthorized
entities to spot element(s) of the task.
Very Low
(1)
Vulnerability in task does not allow any
significant access to unauthorized entities.
A Quantitative Methodology for Security Risk Assessment of Enterprise Business Processes
393
4.2.2 Exploitability
Exploitability is another important attribute of
vulnerability that denotes the ease with which the
vulnerability can be exploited by threat(s)
(Bhattacharjee et al., 2013). It can be determined
with the help of 3 factors as follows:
1) Access Vector (AR). This indicates the type of
access that is required in order to exploit the
vulnerability of a task (say t
j
). It can assume
values on a 3-point scale: (i) if the vulnerability
can be exploited only from within the task t
j
, AR
= 1; (ii) if the vulnerability can be exploited from
a neighbouring task, that is a task (say t
i
) which
has a flow relation with task t
j
(that is t
i
t
j
),
AR = 2; (iii) if the vulnerability can be exploited
from a non-neighbouring task, AR = 3.
2) Attack Complexity (AC). This factor signifies the
amount of difficulty that needs to be encountered
for exploiting vulnerabilities. Higher is the
complexity, lower is the corresponding
exploitability of vulnerability. Different
resources may be required for exploiting
vulnerabilities. Types of resources are: (i)
financial resource; (ii) manpower; (iii)
knowledge or expertise; (iv) tools, techniques,
and infrastructure; and (v) time. The following
scheme is proposed for computing AC: (i) if at
least four of the five resources listed above are
required for launching an attack, then AC = 3;
(ii) if two or three of the resources listed above
are required for launching an attack, then AC =
2; (iii) for all other cases, AC = 1.
3) Authentication Level (AL). This denotes
whether (and how many) authentication is
needed for gaining access to the task element that
contains the vulnerability. It can assume values
on a 3-point scale: (i) if multiple instances of
authentication are required before being granted
access to the task element, then AL = 1; (ii) if a
single instance of authentication is needed, then
AL = 2; (iii) if no authentication is needed, then
AL = 3.
It is obvious from the above discussion that
exploitability of vulnerability is directly proportional
to access vector and authentication level, and
inversely proportional to attack complexity. It can be
computed as:
Exp(v) = ceil((AR * AL) / (3 * AC))
(3)
Here, Exp(v) denotes exploitability of vulnerability
v and ceil denotes ceiling function. While, the
denominator AC is multiplied by 3 to scale down the
value of Exp(v), use of the ceiling function ensures
that the value never reaches 0. Intuitively, if a
vulnerability exists, it is definitely exploitable
(though, the ease of exploitability may vary from
one vulnerability to another); the ceiling function
reflects this notion.
Thus, Exp(v)
(max)
= ceil((3 * 3) / (3 * 1)) = 3
and Exp(v)
(min)
= ceil((1 * 1) / (3 * 3)) = 1
Hence,
Exp(v) ε {1, 2, 3}
(4)
4.3 Threats
Threat (ISO/IEC, 2014) is defined as the potential
cause of an unwanted incident, which may result in
harm to a task (and hence, business process). Threats
exploit vulnerabilities to cause impairment to
elements of a task. Historically, threats have been
categorized based on the type of entity that initiates
the corresponding incident, namely nature (e.g.
earthquake, lightning, etc.), environment (e.g. forest
fire, corrosion, etc.), and human beings (e.g.
malware, information theft, etc.). Such entities are
referred to as primary threat agents; they may, or
may not, need other (secondary) agents to inflict
harm.
After identifying a particular threat, it is
important to compute the probability that the threat
will actually give rise to an incident. A threat might
exist in a passive state without causing any untoward
incident. For example, a volcano may remain
dormant for ages, without causing any harm to its
surrounding regions. On the other hand, cases of
malware infections are on the rise and are causing
serious damage to software and information assets.
The proposed methodology computes threat
probability, referred to as likelihood of occurrence
of threats, as described in the following section.
4.3.1 Likelihood of Occurrence
Likelihood of occurrence (LOC) of threats can be
predicted by combining data about past threat
occurrences with current values of threat parameters
(Bhattacharjee et al., 2014). Specifically, five factors
are considered: (i) past occurrences (p
t
) of threat-
related incidents; (ii) proximity of threat agents to
task elements (a
t
); (iii) existence of motive of an
agent (mvn
t
); (iv) resources to realize a threat (res
t
);
and (v) efficacy of controls implemented (if any) to
mitigate threat (e
c
). Combining these, likelihood of
threat t, for human-induced threats, is obtained as
follows:
LOC(t) = ceil((p
t
+a
t
+mvn
t
+ res
t
) / (2 + e
c
))
(5)
ICISSP 2016 - 2nd International Conference on Information Systems Security and Privacy
394
Divisor 2 is used in Equation 5 to scale down the
value of LOC. For natural and environmental
threats, LOC is given by:
LOC(t) = ceil((p
t
+a
t
+ res
t
) / (2 + e
c
))
(6)
The methods of computation of threat parameters are
as described in (Bhattacharjee et al., 2014). The
computation of a
t
needs some attention, though. In
case of human-induced threats, both physical and
logical access may be crucial. While physical access
is denoted by reachability of threat agents, logical
access is manifested in task authorizations. If a
human threat agent can reach the task element, as
well as has authorization to access it (maybe
acquired illegally), then the value of a
t
is 3. If the
agent can either reach the task element or has
authorization on it, but not both, then the value of a
t
is 2. Finally, if the agent can neither reach the task
element nor has authorization on it, then the value of
a
t
is 1. It has been shown in (Bhattacharjee et al.,
2014) that:
LOC(t) ε {1, 2, 3, 4, 5}
(7)
4.4 Security Concern
After identifying vulnerabilities within task elements
and corresponding threats that can exploit them, the
proposed methodology computes breachability and
security concern values of threat-vulnerability pairs.
Breachability defines the potential of a threat being
able to exploit a given vulnerability (Bhattacharjee
et al., 2013). It is computed as:
B(t, v) = RoundOff(α *LOC(t) + β *Exp(v))
such that (α + β) = 1
(8)
Weights can be customized based on the enterprise
requirements. From Equations 4, 7 and 8, it is
obvious that
B(t, v) ε {1, 2, 3, 4, 5}
(9)
Security Concern (SC) value is computed
considering the breachability and severity of
vulnerability. It denotes the impact that can occur if
vulnerability v is exploited by corresponding threat t
and is given by:
SC(t, v) = ceil(B(t, v) * Sev(v) / 5)
(10)
Since Sev(v)
{1, 2, 3, 4, 5}, it can be seen from
Equations 9 and 10 that:
SC(t, v) ε {1, 2, 3, 4, 5}
(11)
4.5 Risk
Risk is first computed for individual tasks of a
business process (BP). These are then aggregated to
derive risk values for the entire business process.
4.5.1 Risk to Task
After having computed security concern values of all
threat-vulnerability (t-v) pairs for all elements of a
task, they are grouped into three categories –
confidentiality concern (C-concern), integrity
concern (I-concern) and availability concern (A-
concern). The category C-concern contains all those
security concern values that have been derived from
such t-v pairs which can breach the confidentiality
of a task element. Similarly, I-concern and A-
concern categories contain values corresponding to
integrity and availability parameters of a task
element, respectively. If a t-v pair can breach
multiple security parameters, then the pair will
contribute security concern to multiple categories.
For example, the threat “unauthorized data
modification” can exploit the vulnerability “lack of
access control” to breach integrity and availability of
the input element of a task. This gives rise to both I-
concern as well as A-concern for the input element.
The security concern categories for task ti can be
denoted by C-concern(t
i
), I-concern(t
i
) and A-
concern(t
i
). Three separate risk values are obtained
for a task – confidentiality risk, integrity risk and
availability risk. These are computed as follows:
C-risk(t
i
) = ceil(C-value(t
i
) * max(C-concern(t
i
)) / 5)
I-risk(t
i
) = ceil(I-value(t
i
) * max(I-concern(t
i
)) / 5)
A-risk(t
i
) = ceil(A-value(t
i
) * max(A-concern(t
i
)) / 5)
(12)
Here, C-value(t
i
), I-value(t
i
) and A-value(t
i
)
represent the confidentiality, integrity and
availability values of task t
i
as described in Section
4. These values are combined with the maximum
values of C-concern(t
i
), I-concern(t
i
) and A-
concern(t
i
) to obtain confidentiality, integrity and
availability risk values, respectively, of task t
i
. It can
be seen from Equation 11 and Section 4 that
C-risk(t
i
) ε {1, 2, 3, 4, 5}
I-risk(t
i
) ε {1, 2, 3, 4, 5}
A-risk(t
i
) ε {1, 2, 3, 4, 5}
(13)
4.5.2 Risk to Business Process
The next step in the proposed methodology is to
compute the risks to a business process. It has been
shown in Section 3 that a business process consists
of a set of tasks that follow one or more of the
following flow relations: sequential, iterative,
conditional or parallel flow. The process for
combining risk values of tasks is described below.
A Quantitative Methodology for Security Risk Assessment of Enterprise Business Processes
395
1) Sequential Tasks. Risk values of sequential tasks
(Figure 1(a)) are combined by computing their
simple average. This helps in deriving a
reasonable risk value that is neither too high
(which would be the case if the max. risk value
of component tasks was considered), nor too
low, but reflects the actual state of the business
process. One could be tempted to use weighted
average with weights being assigned to tasks
based on their criticality. However, it may be
recalled that during computation of task values,
confidentiality, integrity and availability
requirements of tasks have already been
considered, and hence the same exercise need not
be repeated here. Thus, the combined risk value
for a set of sequential tasks of a business process
will be given by:
C-risk(t
1
,...t
n
) = ceil(Σ C-risk(t
i
) / n)
I-risk(t
1
,...t
n
) = ceil(Σ I-risk(t
i
) / n)
A-risk(t
1
,...t
n
) = ceil(Σ A-risk(t
i
) / n)
(14)
Here, C-risk(t
1
,…t
n
) denotes confidentiality-risk for
sequential tasks t
1
,…t
n
. Similarly, I-risk(t
1
,…t
n
) and
A-risk(t
1
,…t
n
) represent integrity-risk and
availability-risk for t
1
,…t
n
, respectively.
2) Iterative Tasks. A set of iterative tasks (Figure
1(d)) can be viewed as a sequence of tasks whose
risk values may vary between iterations. Hence,
their combined risk value is obtained in the same
manner as for sequential tasks. For example, if
two tasks in sequence, say t
1
and t
2
, are repeated
n times, then it can be assumed that there are
actually 2n tasks, say t
11
, t
21
, t
12
, t
22
,…, where t
11
denotes task t
1
during the first iteration, t
21
denotes task t
2
during the first iteration, t
12
denotes task t
1
during the second iteration, and so
on. Also, their risk values may change over
iterations. Hence, their combined risk values will
be given by:
C-risk(t
1
, t
2
) = ceil(Σ(C-risk(t
1i
) +C-risk(t
2i
)) / n)
I-risk(t
1
, t
2
) = ceil(Σ(I-risk(t
1i
) +I-risk(t
2i
)) / n)
A-risk(t
1
, t
2
) = ceil(Σ(A-risk(t
1i
) +A-risk(t
2i
)) / n),
where i = 1, ..., n
(15)
Here, C-risk(t
1
, t
2
) denotes confidentiality-risk for
iterative tasks t
1
and t
2
. Similarly, I-risk(t
1
, t
2
) and A-
risk(t
1
, t
2
) represent integrity-risk and availability-
risk for t
1
and t
2
, respectively.
3) Conditional Tasks. In cases where one among
several tasks can be executed, depending on the
evaluation result of a condition, the risk value of
the task that is greatest among them, is
considered. For example, in Figure 1(c), one
among tasks T
4
and T
5
will be executed
depending on the result of evaluation of
condition C
1
. Hence, the maximum of the risk
values of T
4
and T
5
will be considered during
computation of risk of the corresponding
business process.
Hence, resultant risk of conditional tasks will be
given by:
C-risk(t
1
,...t
n
) = max(C-risk(t
1
),...,C-risk(t
n
))
I-risk(t
1
,...t
n
) = max(I-risk(t
1
),...,I-risk(t
n
))
A-risk(t
1
,...t
n
) = max(A-risk(t
1
),...,A-risk(t
n
))
(16)
Here, C-risk(t
1
,…t
n
) denotes the resultant
confidentiality-risk for conditional tasks t
1
,…t
n
.
Similarly, I-risk(t
1
,…t
n
) and A-risk(t
1
,…t
n
) represent
resultant integrity-risk and availability-risk for
t
1
,…t
n
, respectively.
4) Parallel Tasks. In cases where several tasks can
be executed simultaneously, the resultant risk
value is obtained as a simple average of the risk
values to those tasks. For example, in Figure
1(b), tasks T
9
and T
10
will be executed in
parallel. Hence, the simple average of risk values
of T
9
and T
10
will be considered during
computation of risk of the corresponding
business process.
Hence, resultant risk of parallel tasks will be given
by:
C-risk(t
1
,...t
n
) = ceil(Σ(C-risk(t
i
) / n)
I-risk(t
1
,...t
n
) = ceil(Σ(I-risk(t
i
) / n)
A-risk(t
1
,...t
n
) = ceil(Σ(A-risk(t
i
) / n),
where i = 1, ..., n
(17)
Here, C-risk(t
1
,…t
n
) denotes confidentiality-risk for
parallel tasks t
1
,…t
n
. Similarly, I-risk(t
1
,…t
n
) and A-
risk(t
1
,…t
n
) represent integrity-risk and availability-
risk for t
1
,…t
n
, respectively.
As shown in Figure 2, a business process can be
a combination of various kinds of task structures.
The composite risk values of the entire business
process can be obtained as follows:
Replace individual sets of parallel tasks with a
single task having risk value as stated in
Equation 17;
Replace individual sets of conditional tasks with
a single task having risk value as stated in
Equation 16;
Expand sets of iterative tasks to form a simple
sequence of tasks having individual risk values;
Finally, compute the risk values of the business
process, which now contains a single sequence of
tasks, using Equation 14.
The risk values of a business process are denoted by
C-risk(BP
j
), I-risk(BP
j
) and A-risk(BP
j
), which can
ICISSP 2016 - 2nd International Conference on Information Systems Security and Privacy
396
be interpreted as confidentiality-, integrity- and
availability-risk of business process BP
j
.
5 DISCUSSION
The proposed methodology computes security risks
to business processes of an enterprise. Risks that
have the potential to breach confidentiality, integrity
and availability of business processes are analyzed at
individual task level and their values are calculated.
These values are then aggregated to derive
consolidated risk values for corresponding business
processes. Thus, three separate risk values, namely
confidentiality risk, integrity risk and availability
risk, are computed for each business process. It has
been assumed that maintenance of correct order of
the tasks of a business process, denoted by the flow
relation , is the responsibility of an enterprise. As
discussed in Section 3, improper configuration of the
flow relation can lead to disastrous consequences.
Also, change in the flow relation actually gives rise
to a business process that is different from the
desired one as is obvious from the definition of
business process given by Equation 2. Hence, in
such a scenario, the risk that would be computed
would, in fact, pertain to the new (incorrectly
configured) business process.
This risk assessment methodology will help an
enterprise to decide on risk mitigation strategies that
can protect its more critical security parameters.
Owing to the generic nature of the methodology,
both external and internal risks can be addressed.
Besides, the steps of risk computation can be re-
traced to identify the tasks, and their specific
elements, that are the biggest contributors to the risk
values. Another approach could be to maintain
records of the threat-vulnerability pairs, their values
and the corresponding security parameters that they
can breach. Such an exercise can be easily
implemented by following the steps of the proposed
methodology as has been described in earlier
sections. This would help in designing a better risk
mitigation technique that can be effective in
addressing the concerns of the enterprise.
It may be noted that the proposed methodology
provides a technique for evaluating business
processes, vulnerabilities and threats, and combining
them to derive risk values. Major categories of
vulnerabilities have also been described in Section
4.2. However, specific vulnerabilities and threats
vary between enterprises and enterprise-sectors.
Hence, actual identification of these risk factors is
the task of the operational team that is involved in
the computation of business risks.
A detailed case study that demonstrates the steps
of the proposed methodology is included in the
following section.
6 CASE STUDY
The case study refers the Loan Management Process
shown in Figure 2. It can be seen that tasks t
1
to t
8
follow a sequential flow wherein completion of t
1
is
a pre-requisite for initiation of t
2
, and so on.
Similarly, t
8
is a pre-requisite for t
9
and t
10
which, in
turn, lead to task t
11
. Finally, execution of t
4
or t
5
or
t
11
should be completed before execution of task t
12
.
The following assumptions may be made: task t
1
requires infrastructure, money and time, while tasks
t
3
to t
12
require all resources stated in Section 3.
Besides, all the tasks require some input and
generate some outcome.
Considering the pre-requisites, resources, input,
and outcome of tasks and guidelines provided in
Table 1, Table 2 and Table 3, appropriate C, I and A
values can be assigned to the tasks of loan
management process as shown in Table 5.
Table 5: C, I and A values of the tasks of loan
management process.
Task_ID C-value I-value A-value
t
1
4 4 4
t
2
3 3 4
t
3
3 3 4
t
4
3 4 4
t
5
3 4 4
t
6
4 4 4
t
7
4 4 4
t
8
4 4 4
t
9
3 4 4
t
10
3 4 4
t
11
3 4 4
t
12
3 4 4
Table 6: Vulnerabilities in the tasks of loan management
process.
Task_ID Vulnerability Sev Exp
t
1
, t
9
, t
10
,
t
11
Improperly configured functions may
generate erroneous outcome, or may not
produce any outcome (v
1
)
3 2
t
4
, t
5
Lack of training of concerned personnel
may lead to incorrect execution of tasks
(v
2
)
4 2
t
6
, t
7
, t
8
, t
9
, t
10
Improperly configured infrastructure
may hinder task execution (v
3
)
3 2
t
11
, t
8
, t
9
, t
10
Lack of output validation may cause
leakage of confidential information to
unauthorized entities (v
4
)
3 2
A Quantitative Methodology for Security Risk Assessment of Enterprise Business Processes
397
Threats that can exploit the above vulnerabilities
are listed in Table 7. LOC values of threats are
computed applying Equations 5 and 6.
Applying Equation 8 and assuming α = β = 0.5,
B(th
1
, v
1
) = 0.5 * 4 + 0.5 * 2 = 3.
Similarly, B(th
2
, v
2
) = 3, B(th
3
, v
3
) = 3, and
B(th
3
, v
4
) = 3.
Applying Equation 10, SC(th
1
, v
1
) = (3 * 3) / 5 =
2.
Table 7: Vulnerabilities and threats.
Vulnerability Threat LOC
v
1
Error in use (th
1
) 4
v
2
Corruption of data (th
2
) 3
v
3
Illegal processing (th
3
) 3
v
4
Illegal processing (th
3
) 3
Similarly, SC(th
2
, v
2
) = 3, SC(th
3
, v
3
) = 2, and
SC(th
3
, v
4
) = 2.
It may be seen that threat th
1
can exploit
vulnerability v
1
to breach the availability of tasks t
1
,
t
9
, t
10
and
t
11
. So, applying Equation 12,
A-risk(t
1
) = (4 * 2) / 5 = 2, A-risk(t
9
) = 2, A-
risk(t
10
) = 2, and A-risk(t
11
) = 2.
Threat th
2
can exploit vulnerability v
2
to breach
the availability of tasks t
4
and
t
5
. Hence, A-risk(t
4
) =
(4 * 3) / 5 = 3; A-risk(t
5
) = (4 * 3) / 5 = 3.
Threat th
3
can exploit vulnerabilities v
3
and v
4
to
breach the availability of tasks t
6
, t
7
and t
8
. So, A-
risk(t
6
) = 2, A-risk(t
7
) = 2, and A-risk(t
8
) = 2.
Applying Equation 14, combined risk of
sequential tasks t
6
, t
7
and
t
8
= 2.
Applying Equation 16, resultant risk of
conditional tasks t
4
and
t
5
= 3.
Applying Equation 17, resultant risk of parallel
tasks t
9
and t
10
= 2.
Assuming that tasks t
2
, t
3
and t
12
do not have any
identified risks, the composite risk value of the loan
management process = (2+2+3+2+2) / 5 = 3.
7 CONCLUSION AND FUTURE
WORK
A methodology for assessing security risks of
enterprise business processes has been presented in
this paper. The paper begins by stating a formal
definition of business process based on tasks and
their flow relation. The types of flow relations,
namely sequential, parallel, conditional and iterative,
have been clearly illustrated. The component
elements of a task and their functionalities have been
detailed. The manner in which a task can be related
to another task has also been described. Such
relation is defined either by the dependence of a task
on the outcome of another task, or due to triggering
of one task by another. A technique for deriving the
confidentiality, integrity and availability
requirements of tasks has been proposed. The paper
also models the vulnerabilities within, and threats to,
task elements, along with their attributes. These
include severity and exploitability of vulnerabilities,
and likelihood of occurrence of threats. Finally,
procedures for computing security concern and risk
values of tasks and business processes have been
detailed. The paper concludes with a case study that
illustrates the proposed methodology in its entirety.
It is hoped that the technique presented in this
paper will help implementers to correctly model, and
identify and assess critical risks to, business
processes. This will, in turn, enable the design and
implementation of an effective protection
mechanism for enterprise business processes.
Future work is geared towards the development
of a tool based on the proposed methodology.
Results of risk assessment using the tool will help in
validating the technique that will lead to further
improvement of the mechanism. Another interesting
extension could be the formulation of techniques for
combining business process risks and asset-based
risks to derive the overall value of enterprise risk.
This could be useful in the preparation of
dashboards for viewing by top management and
making critical investment decisions.
ACKNOWLEDGEMENTS
This research was partially supported by grants
allocated by the Department of Electronics and
Information Technology, Govt. of India.
REFERENCES
Armando, A., Ponta, S. E., 2011. Model Checking of
Security-Sensitive Business. In CRiSIS’11, 6
th
International Conference on Risks and Security of
Internet and Systems. IEEE, pp. 66-80.
Bhattacharjee, J., Sengupta, A., Mazumdar, C., 2013. A
Formal Methodology for Enterprise Information
Security Risk Assessment. In CRiSIS’13, 8
th
International Conference on Risks and Security of
Internet and Systems. IEEE, pp. 1-9.
Bhattacharjee, J., Sengupta, A., Mazumdar, C., 2014. A
Formal Methodology for Modeling Threats to
Enterprise Assets. In ICISS’14, 10
th
International
Conference on Information Systems Security. Springer,
pp. 149-166.
ICISSP 2016 - 2nd International Conference on Information Systems Security and Privacy
398
Denning, D. E., 1976. A Lattice Model of Secure
Information Flow. Communications of ACM, ACM.
Vol. 19, No. 5, pp. 236-243.
ENISA. 2015. Ebios. Available at:
https://www.enisa.europa.eu/activities/risk-
management/current-risk/risk-management-
inventory/rm-ra-tools/t_ebios.html (Accessed: 5 June
2015).
ISO/IEC JTC 1 IT SC 27, 2011. Information technology –
Security techniques - Information security risk
management, ISO/IEC 27005:2011, ISO/IEC. Geneva,
2
nd
Edition.
ISO/IEC JTC 1 IT SC 27, 2014. Information technology –
Security techniques - Information security
management systems – Overview and vocabulary,
ISO/IEC 27000:2014, ISO/IEC. Geneva, 3
rd
Edition.
Jakoubi, S, Tjoa, S., Goluch, S., Kitzler, G., 2010. Risk-
Aware Business Process Management—Establishing
the Link Between Business and Security. Book of
Complex Intelligent Systems and Their Applications,
Springer Science+Business Media, LLC. Vol. 41, pp.
109-135.
Khanmohammadi, K., Houmb, S. H., 2010. Business
Process-based Information Security Risk Assessment.
In 4
th
International Conference on Network and
System Security, IEEE, pp. 199-206.
Lowis, L., Accorsi, R., 2011. Vulnerability Analysis in
SOA-Based Business Processes. IEEE Transactions
On Services Computing, IEEE. Vol. 4, No. 3, pp. 230-
242.
Marchesini, S., Viganò, L., 2011. A Hierarchy of
Knowledge for the Formal Analysis of Security-
Sensitive Business Processes. In CRiSIS’11, 6
th
International Conference on Risks and Security of
Internet and Systems. IEEE, pp. 1-10.
MEHARI, 2010. Risk analysis and treatment Guide.
Available at:
https://www.clusif.asso.fr/fr/production/ouvrages/pdf/
MEHARI-2010-Risk-Analysis-and-Treatment-
Guide.pdf (Downloaded: 10 August 2015).
OMG, 2011. Business Process Model and Notation,
BPMN, Object Management Group. Massachusetts,
2
nd
edition.
Peltier, T. R., 2010. Information Security Risk Analysis,
CRC Press. Florida, 3
rd
Edition.
Soanes, C., Stevenson, A. (ed.), 2011. Concise Oxford
English Dictionary, Oxford University Press. New
York, 12
th
Edition.
Sun, S. X., Zhao, J. L., Nunamaker, J. F., Sheng, O. R. L.,
2006. Formulating the Data-Flow Perspective for
Business Process Management. Information Systems
Research, INFORMS. Vol. 17, No. 4, pp. 374-391.
Tjoa, S., et al., 2011. A Formal Approach Enabling Risk-
Aware Business Process Modeling and Simulation.
IEEE Transactions On Services Computing, IEEE.
Vol. 4, No. 2, pp. 153-166.
A Quantitative Methodology for Security Risk Assessment of Enterprise Business Processes
399