The challenges of so s assurance

Post on 01-Nov-2014

176 views 0 download

description

 

Transcript of The challenges of so s assurance

© The University of York© The University of York

The Challenges of SoS Assurance:

A Safety Case Perspective

George Despotou

© The University of York

Overview

• Systems of Systems

– Example

• Safety lifecycle

• (Top down) safety case

• SoS hazards and related arguments

• Overview of the safety case

• Related challenges

© The University of York

Systems of Systems

• Distinct class of systems

demonstrating certain attributes

• Geographical dispersion

• Overall objectives

• Complexity

• Multiple elements

• Autonomy

• Collaboration

• Communication

• Heterogeneity

• Common methods often

fall short for these type of

systems

© The University of York

Challenges in SoS Analysis

• Evolving systems

– Constantly evolving and adapting

• Emergent behaviour

– Behaviour affected by this of its components

• Functional composition

– SoS functions composed of elements’ functions

• Network centred

– Heavy reliance on communications

• Multi-attribute failures

– Element failures propagate and manifest as different types

– Trade-offs

© The University of York

SoS Safety

Lifecycle

• Extends ARP

lifecycle

• Incorporates SoS

modelling

approaches

(MODAF/DODAF)

ABM Stages Preliminary

Scenario-based

Hazard

Identification

Activity-based

Hazard

Analysis

Preliminary

SoS Safety

Assessment

Co

mm

on

Mo

de

& E

mer

gen

t (S

oS

) H

aza

rds

Overall

Scenario and

Capabilities

Activities,

Actors and

Information

System

Functions,

Data and

connectivity

System

Realisation and

Infrastructure

System

Safety

Assessment

often black box

SoS Safety

Assessment

Safety Synthesis

© The University of York

SoS Safety Case

• Safety case principles are the same

• Argument: series of statements (claim) about properties (safety) of the system

• Evidence supporting the claims

SoS Safety Argument

Evidence

© The University of York

Top down start

• Definition of safety always

requires

consideration of operation

(relevance)

• Capability based in SoS

• Scope not always

clear

SoS Safety Argument

Evidence Evidence

Elaboration

Definition of Safety

(argument)

© The University of York

Definition of safety context

• Scenario based approaches

• Different acceptability criteria for each

• Safety tolerance (risk)

• Operational fitness

– Performance, security,

availability etc.

• Identification of stakeholders goals

MODAF Ov1

•Hidden sub-scenarios may be

present (roles)•e.g. S&R, intelligence, suppression, maintenance etc.

•Understanding the variations

of requirements in each•Crucial for later stages (ALARP)

© The University of York

How is the capability offered?

• Many systems

collaborating

• Different vendors

• Different technologies, standards, methods

• Often in isolation

(competition?) with

each other

• However, common

operator/customer

– Customer structure

Supplier A Supplier B Supplier C

Supplier D Supplier E

Army

Air Force

Comms

Command

Allies Opera

tor

Facets

© The University of York

Multiple supplier safety caseDefinition of Safety

(argument)

Argument for

System

X

Argument for

System

Y

Argument for

System

Z

Evidence for

system X

Evidence for

system Y

Evidence for

system Z

Assura

nce C

ase f

or

Sys

tem

X

Assura

nce C

ase f

or

Sys

tem

Y

Assura

nce C

ase f

or

Sys

tem

Z

Supplier A Supplier B Supplier C

© The University of York

Multiple supplier safety caseDefinition of Safety

(argument)

Argument for

System

X

Argument for

System

Y

Argument for

System

Z

Evidence for

system X

Evidence for

system Y

Evidence for

system Z

Assura

nce C

ase f

or

Sys

tem

X

Assura

nce C

ase f

or

Sys

tem

Y

Assura

nce C

ase f

or

Sys

tem

Z

Supplier A Supplier B Supplier C

This is a

challenging step

© The University of York

SoS Hazard Anlysis

• Focusing on individual systems is a starting point (e.g.

DDA)

– But not sufficient

• Obscure (complex) causal chain to hazards

– Combination, propagation and transformation of failures

• The degree of this phenomenon can be profound

– Controls for a single hazard may need to be implemented by multiple suppliers

– Often difficult to predict behaviour (e.g. autonomy)

• Not always deviant behaviour

– An SoS can operate as designed

• Yet resulting in (emergent) hazards

© The University of York

The role of simulation

• Exhaustive

– Cannot claim that

– Only to a degree of confidence

• (Exhaustive) manual analysis almost impossible

– Scale

– Configurations

• Known and unknown (adaptability)

• Simulation offers a ‘brute force’ approach to identifying hazardous conditions

– Fault injection

© The University of York

Challenges for simulation

• Two well known challenges

– Valid

• Do the simulated models represent the right system

– Verified

• Has the simulation been implemented correctly?

• Both may undermine our confidence in simulation

• Effectiveness of simulation is unknown

– Cannot argue whether the potential risk reduction from

simulation justifies the cost

• Need for more empirical evidence about simulation

• And tactics to argue/achieve validation and verification

© The University of York

Safety Requirements

• System AC will support the safety argument

– i.e. hazard oriented argument

• Follows identification of SoShazards

– And the contribution of each system

to them

• Assumes ‘straight forward’

contribution of system operation to

hazards

– Discharging safety requirements to

suppliers

Ass

ura

nce

Ca

se X

Definition of Safety

(argument)

Ass

ura

nce

Ca

se Y

Ass

ura

nce

Ca

se Z

Dependency Definition

(hazard controls argument)

© The University of York

(Addressing) Cross supplier

hazards

• Hazards occurring from

combination of failures

• Addressing the hazard may

result in depending on another

suppliers assurance case

– Challenges during acquisition

• Cannot anticipate specific

dependencies in advance

– Impact on acquisition process

(and contracts)

Assurance

Case X

Definition of Safety

(argument)

Dependency Definition

(hazard controls argument)

Assurance

Case Y

Evidence

Evidence

© The University of York

Hazard controls

• Discharging (safety related) requirements

– Potential dependencies among suppliers

• Coordination throughout acquisition is crucial

• Safety policy

– Implemented on top of the SoS elements

– Who will be responsible to monitor/enforce it

• Controller vs. distributed policy

• Any additional requirements to suppliers?

– Onus on the operator to analyse hazards and argue safety

• May result in the operator having to produce evidence

© The University of York

SoS Safety

Case

Definition of Safety

(argument)

Hazard Controls Argument

Ass

ura

nce

Ca

se X

Opera

tional S

C X

Ass

ura

nce

Ca

se Y

Opera

tional S

C Y

Ass

ura

nce

Ca

se Z

Opera

tional S

C Z

SoS Safety

Policy

Argument

SDR

Argument

Other means

(e.g.

controller)

AC

ab

ou

t oth

er

safe

ty r

elate

d

syst

ems

© The University of York

Safety Case - Responsibility

• Not clear

– Except for assurance cases for individual systems

– Interweaved interest

• Supplier needs to be aware of hazard controls

– Dependencies on their system

» SDR and policy

– Claims made, which may depend on other supplier

• Who provides the contextual information?

– E.g. SoS hazard analysis

• operator: ability to do so?

• supplier: access to information?

– Inevitable assumptions

© The University of York

SoS Acquisition Challenges

• Increasingly apparent that collaboration amongst stakeholders is inevitable

• Commercial/competition issues

• Customer/operator driven process

– Clear identification of stakeholder interfaces

– Cannot be a single demonstration

• Continuous process and support

• Not zero sum (?)

– It is in the interests of the suppliers to increase

collaboration (?)

© The University of York

Synopsis

• SoS demonstrate a distinct combination of

characteristics

• Challenged to safety analysis

• Safety case principles are the same

• SoS level hazards are difficult to identify

• Allocation of requirements can involve multiple

stakeholders

• Often, dependencies between suppliers

• Responsibility of the safety case is unclear

• Some of these issues can be incorporated in the

acquisition process

© The University of York

Further Info

• SoS community

– www.dependablesos.org

• Informal; contributions welcome

• www-users.cs.york.ac.uk/~george

• Assurance cases

– GSN: www.goalstructuringnotation.info

– AC editor: http://code.google.com/p/acedit/