PPT

advertisement

VERIFICATION OF

PARAMETERIZED SYSTEMS

MONOTONIC ABSTRACTION IN

PARAMETERIZED SYSTEMS

Parosh Aziz Abdullah, Giorgio Delzanno, Ahmed Rezine

NAVNEETA NAVEEN PATHAK

AGENDA

INTRODUCTION

PARAMETERIZED SYSTEMS

 TRANSITION SYSTEMS

 ORDERING

 MONOTONIC ABSTRACTION

Monotonic Abstraction in Parameterized Systems 2

INTRODUCTION

 Monotonic Abstraction as a simple and effective method to prove safety properties for

Parameterized Systems with linear topologies.

 Main idea : Monotonic Abstraction for considering a transition relation that is an overapproximation of the one induced by the parameterized system.

Monotonic Abstraction in Parameterized Systems 3

MODEL CHECKING + ABSTRACTION

Infinite-State

System

Abstraction

Model

Checking

Finite-State

System

Monotonic Abstraction in Parameterized Systems 4

AGENDA

INTRODUCTION

PARAMETERIZED SYSTEMS

 TRANSITION SYSTEMS

 ORDERING

 MONOTONIC ABSTRACTION

Monotonic Abstraction in Parameterized Systems 5

PARAMETERIZED SYSTEMS

P

1

P

2

P

3

..........

P

N

P

2

.........

P

1

P

4

P

N

P

3

.........

 AIM : To verify correctness of the systems for the whole family of Parameterized Systems.

Monotonic Abstraction in Parameterized Systems 6

DEFINITION

A parameterized system P is a triple (Q,X, T ),

Q - set of local states,

X - set of local variables,

T - set of transition rules.

A transition rule t is of the form: t: [ q | grd → stmt | q´ ] where q, q´ ϵ Q grd → stmt is a guarded command grd ϵ B(X) U G(X U Q) stmt : set of assignments

Monotonic Abstraction in Parameterized Systems 7

A process moves from

Idle to Black state when it wants to access its critical section.

Parameterized System, P = (Q,T)

Q = {Green, Black, Blue, Red} and T = {t where t t

2, t

1,

3 t

5, t

4 t

6

– Existential Rule

Idle State – Initially all

– Universal Rules state t

2, t

3.

– Local transition rules processes are in this t

4, t

5,

Once a process moves from Black to Blue state, it

“closes the door” on all processes in

Idle state

V LR t

2 t

1 t

6 t

5 t

6

}

Critical State

Eventually a process will enter this state

∃ L t

3 t

V L

4

Monotonic Abstraction in Parameterized Systems 8

AGENDA

INTRODUCTION

PARAMETERIZED SYSTEMS

 TRANSITION SYSTEMS

 ORDERING

 MONOTONIC ABSTRACTION

Monotonic Abstraction in Parameterized Systems 9

TRANSITION SYSTEMS

A transition system T is a pair (C, ⇒ ) where,

C - (infinite) set of configurations ,

⇒ - binary relation on C,

⇒ * - reflexive transitive closure of ⇒

A configuration c ϵ C is a sequence u

1 states.

, ...... , u n of process i.e. corresponding to an instance of the system with n processes.

Monotonic Abstraction in Parameterized Systems 10

The word below represents a configuration in an instance of system with 5 processes.

Valid Transitions t

3

Invalid Transitions t

3

Monotonic Abstraction in Parameterized Systems 11

Initial Configuration

Bad Configuration

All configurations that have atleast 2 RED processes

AIM : Init * Bad ?

Monotonic Abstraction in Parameterized Systems 12

AGENDA

INTRODUCTION

PARAMETERIZED SYSTEMS

 TRANSITION SYSTEMS

 ORDERING

 MONOTONIC ABSTRACTION

Monotonic Abstraction in Parameterized Systems 13

ORDERING

c

1 c

1

, c

2

≤ c

2

– configurations

- c

1 is a subword of c

2 e.g. ≤

Upward Closed Configurations

Set U of configurations is upward closed, if whenever c ϵ U and c ≤ c ´ then c ´ϵ U.

c – configuration, ĉ – denotes upward closed set U:= {c ´ | c ≤ c´} ĉ contains all configurations larger than c w.r.t. ordering ≤.

i.e. c is the generator of U

Monotonic Abstraction in Parameterized Systems 14

Why Upward Closed Sets ?

1. All sets of Bad configurations (which are worked upon) are upward closed.

2. Upward closed sets have an efficient symbolic representation.

i.e. For an upward closed set U, there are configurations c

1

, ..... , c n with U = ĉ

1

U......U

ĉ n

Monotonic Abstraction in Parameterized Systems 15

Coverability Problem for Parameterized

Systems

To analyze safety properties.

PAR-COV

Instance

• Parameterized System, P = (Q,X,T)

• C

F

– upward-closed set of configurations

Question

Init * C

F

?

Monotonic Abstraction in Parameterized Systems 16

Backward Reachability Analysis

For a set of configurations, C

Use Pre(C) := {c | ∃c´ ϵ C; c → c´ }

IDEA : i.

Start with set of bad upward-closed configurations.

ii. Apply function Pre repeatedly generating sequence U

0

U

2

,.... where

,

U

1

0

, U

:= Bad, and U i+1

:= U i

+ Pre(U i

) for all i ≥ 0

Observation : set U i characterizes set of configurations from which set Bad is reachable within i steps

Monotonic Abstraction in Parameterized Systems 17

MONOTONICITY

Monotonicity implies that upward closedness is preserved through the application of Pre.

Consider:

U – upward closed set, c

1

– member of Pre(U) and c

2

≥ c

1

By Monotonicity, it can be proved that c

2 is also a member of Pre(U)

Monotonic Abstraction in Parameterized Systems 18

AGENDA

INTRODUCTION

PARAMETERIZED SYSTEMS

 TRANSITION SYSTEMS

 ORDERING

 MONOTONIC ABSTRACTION

Monotonic Abstraction in Parameterized Systems 19

MONOTONIC ABSTRACTION

An abstraction that generates over-approximation of the transition systems.

The abstract transition system is monotonic.

Hence, allowing one to work with upward closed sets.

c

1

≥ c

1

´

A c

2

Monotonic Abstraction in Parameterized Systems 20

Local transitions are monotonic!

Consider the local transition, c

1 t

2

= = c

3

Configuration c

2

= c

2

= t

2

This leads to c

4

≥ c

2 and also maintains c

3

≤ c

4

.

Monotonic Abstraction in Parameterized Systems c

4

21

Existential transitions are monotonic!

Consider the existential transition: c

1

= t

3

Configuration, c

2

= c

2

= t

3

= c

3

= c

4

Leading to c

4

≥ c

3

Monotonic Abstraction in Parameterized Systems 22

Non-monotonicity of Universal transitions

Consider the following Universal transition: c

1

= t

4 = c

3 t

4 can be applied to c

1 as all process in the left context of the active process satisfy the condition of transition.

Now consider c

2 c

1

≤ c

But t

4

2

= is not enabled from c

2 since the left context of the active process violates the conditions of transition.

Monotonic Abstraction in Parameterized Systems 23

Solution!

1. Work with Abstract transition relation →

A

.

2.

A is an monotonic abstraction (over-approximation) of the concrete relation → .

3. When t is universal, we have: c

1

A c

2 iff c

1

´ → c

2 for some c

1

´ ≤ c

1 i.e. t

A

Since

≤ t →

Monotonic Abstraction in Parameterized Systems 24

Solution.....

Since,

 c

1

 c

1

≤ c

2

A c

3 implies c

2

A c

3

Hence, Abstract transition relation is Monotonic, w.r.t.

Universal Transitions.

The Abstract transition relation is and over-approximation of the original transition relation

↓↓

If a safety property holds in the abstract model, then it will also hold in the concrete model.

Monotonic Abstraction in Parameterized Systems 25

Coverability Problem for Approximate

Systems

APRX-PAR-COV

Instance

• Parameterized System, P = (Q,X,T)

• C

F

– upward-closed set of configurations

Question

Init *

A

C

F

?

Monotonic Abstraction in Parameterized Systems 26

A

= ( U

1

)

1 reflects the approximation of universal quantifiers

Since ⊆

A

A negative answer to APRX-PAR-COV implies a negative answer to PAR-COV.

Monotonic Abstraction in Parameterized Systems 27

CONCLUSION

Monotonic Abstraction in Parameterized Systems 28

 Introduction to our topic.

 Overview of Parameterized Systems using a simple example.

 (Infinite) Transition Systems arising from parameterized systems.

 Introduced Ordering on the set of configurations.

 Definiton and explanation of Monotomic Abstraction; based on the parameterized systems example.

Monotonic Abstraction in Parameterized Systems 29

Thank you for your attention.

Monotonic Abstraction in Parameterized Systems 30

Download