D.3.3 ALGORITHMS FOR INCREMENTAL ... - SecureChange

10 F. Massacci and L.M.S. Tran
[5] Data exchanged
12%
42%
[12] Basic data
exchanged with
DMAN
DMAN
[13] Advanced
data exchanged
with DMAN
12% 42% 46%
[5] Data exchanged
[9] Data
exchanged with
adjacent sectors
[5] Data exchanged
[5] Data
exchanged
[5] Data
exchanged
[9] Data
exchanged with
adjacent sectors
[5] Data exchanged
[12] Basic data
[13] Advanced
exchanged with
data exchanged
DMAN
with DMAN
[9] Data
exchanged with
adjacent sectors
[12] Basic data
exchanged with
DMAN
[9] Data
exchanged with
adjacent sectors
[13] Advance data
exchanged with
DMAN
[12] Basic data
exchanged with
DMAN
[9] Data
exchanged with
adjacent sectors
[13] Advance data
exchanged with
DMAN
46%
[9] Data
exchanged with
adjacent sectors
(a) Tree-like representation
(b) Swimlane-based representation
Fig. 3 Visualization of evolution rules.
The tree-like representation, Fig. 3(a), basically is a tree-like directed graph,
where a node represent a subpart in the goal model. The directional connection
between two nodes means the source node evolves to target node. In other words,
the source node is the original model (or subpart), and target node is the evolved
model (subpart). A connection is labeled with an evolution probability p i . In two
validation sessions with ATM experts the tree-like model was found more intuitive
and we have therefore decided to implement it in our modelling CASE tool.
5 Game-Theoretic Approach Accounting for Evolution Probability
Our notion of observable rules includes an associated probability. Probabilities are
often taken for granted by engineer and scientists but in our setting it is important
to understand the exact semantics of this notion.
Basically, there are two broad categories of probability interpretation, called
“physical” and “evidential” probabilities. Physical probabilities, in which frequentist
is a representative, are associated with a random process. Evidential probability,
also called Bayesian probability (or subjectivist probability), are considered to
be degrees of belief, defined in terms of disposition to gamble at certain odds; no
random process is involved in this interpretation.
To account for probability associated with an observable rule, we can use the
Bayesian probability as an alternative to the frequentist because we have no event
to be repeated, no random variable to be sampled, no issue about measurability
(the system that designers are going to build is often unique in some respects).
However, we need a method to calculate the value of probability as well as to
explain the semantic of the number. Since probability is acquired from the requirements/objectives
eliciting process involving the Domain Experts, we propose
using the game-theoretic method in which we treat probability as a price. It seems
to be easier for Domain Experts to reason on price (or cost) rather than probability.
The game-theoretic approach, discussed by Shafer et al. [35] in Computational
Finance, begins with a game of three players, i.e. Forecaster, Skeptic, and Real-