Synthesis of Behavioral Models from Scenarios

1
Synthesis of Behavioral Models from Scenarios

• Sebastian Uchitel, Jeff Kramer, Jeff Magee
• IEEE Transactions on Software Engineering
• February 2003

June 10, 2003 Presented by, Younju Oh
2
Contents

• Introduction
• Background
• MSC Semantics
• Synthesis of Behavioral Models
• Analysis
• Case Study
• Conclusion

3
Introduction (1/3)

• Scenario-based specifications
• Useful in requirements specification
• Proven to be good for bridging gap between
  stakeholders and developers
• Scenarios
• Describe how system components,environment, users
  interact
• One scenario partial behavior
• Widespread notation is MSC

4
Introduction (2/3)

• Why Scenario synthesis ?
• Partial behavior - relatively little information
• Scenarios may be provided by different
  stakeholders
• Address different system functionality
• Combination More complete system description
• Relations between scenarios

5
Introduction (3/3)

• Papers Approach

Scenario-based specification, Synthesis, Behavior
Analysis
Finite Sequential Processes (FSP)
Labeled Transition System (LTS)
6
Message Sequence Charts (MSC) syntax
bMSC

• bMSCs - describes finite interactions between a
  set of components
• hMSCs - directed graphs, nodes are bMSCs or
  hMSCs, edges indicate possible continuations

7
Multiple Scenarios (1/2)

• Relation between scenarios
• Scenario Composition Using graph-like notation
  (hMSC)
• Scenario reuse
• Large number of short scenarios ? composed in
  complex ways
• State Identification identify common states in
  different scenarios
• No need to split into small parts
• Requires stakeholders to to think of system in
  states

Scenario composition State identification
8
Multiple Scenarios (2/2)
9
ATM system
10
MSC Semantics

• Labeled Transition Systems (LTS)

11
Labeled Transition System
Minimized Component LTS for Consortium

• Components defined by MSC specification
• Result of putting together instance LTSs and
  their continuation relation
• Semantics of MSC specifications
• LTSs and parallel composition

12
Synthesis of Behavioral Models

• LTS model can be synthesized from MCS
  specification
• MSC ? FSP ? LTS
• Synthesis Algorithm
• Translates MSC specification to FSP
• Labeled Transition System Analyzer (LTSA)
• Verification tool for concurrent systems
• Builds LTS model from FSP
• Check that specification satisfies behavioral
  properties

13
Synthesis Algorithm (1/3)
Bad Bank Account
User
ATM
Consortium
Bank
Enter Password
Verify Account
Verify card with bank
Verifying With Bank
Waiting
Verifying
Checking
Bad bank account
Bad Account
14
Synthesis Algorithm (2/3)
4
ATM
(Init, B_CustomerArrives) (E_BadBankPassword,
B_UserCancel1) (E_CustomerArrives, B_UserCancel1)
(E_BadBankPassword, B_BadBankAccount)
(E_CustomerArrives, B_BadBankAccount)
(E_BadBankPassword, B_UserCancel2)
(E_CustomerArrives, B_UserCancel2)
(E_BadBankPassword, B_ BadBankPassword)
(E_CustomerArrives, B_ BadBankPassword)
(E_UserCancel2, B_CustomerArrives)
(E_UserCancel1, B_ CustomerArrives)
(E_BadBankAccount, B_CustomerArrives)
ATM
Verifying
E_CustomerArrives
Bad Account
Enter Password
Verify Account
Verifying
B_CustomerArrives
ATM
ATM
Verifying
B_BadBankAccount
Bad Account
Enter Password
Verify Account
Verifying
E_BadBankAccount
15
Synthesis Algorithm (3/3)
ATM
5
E_CustomerArrives
Enter Password
E_CustomerArrives (EnterPassword -
verifyAccount - Verifying)
Verify Account
Verifying
Multiple Definitions of local processes
E_CustomerArrives (cancel - canceledMessage -
ejectCard - requestTakeCard - takeCard - Init
enterpassword - verifyAccount -
Verifying)
16
Analysis

• FSP ? LTS
• Synthesized model can be analyzed to provide
  useful feedback
• LTSA
• Applies model checking algorithms that traverse
  the system LTS verifying properties
• LTSA provides automated rigorous analysis of
  behavioral models
• Other representations of large systems does not
  favor rigorous analysis
• LTSA minimizes models

17
Case Study (1/3)

• Whittle, Schumann
• Generate UML Statecharts from scenarios and pre-,
  post-conditions in Object Constraint Language
  (OCL)
• No state labeling or hMSC
• Proposed approach
• LTS from MSC and OCL

18
Case Study (2/3)
1. Scenarios describe system behavior from the
start
19
Case Study (3/3)
20
Conclusion

• Labeled transition systems and parallel
  compositions
• ? define semantics of MSC language
• Synthesis algorithm generates FSP
• From FSP, LTS behavior model is built for
  analysis in LTSA
• hMSC, state labels promote scenario reuse,
  explicit info
• LTS Behavior analysis in LTSA
• Approach provides basis for common approach to
  other approaches