a logical framework for anticipation of network · a logical framework for anticipation of network...

34
Elie Bursztein [email protected]cachan.fr A Logical Framework for Evaluating Network Resilience Against Faults and Attacks A Logical Framework for Anticipation of Network A Logical Framework for Anticipation of Network Incidents Elie Bursztein and Jean GoubaultLarrecq Phd Student LSV ENSCACHAN CNRS INRIA DGA Phd Student LSV ENSCACHAN CNRS INRIA DGA

Upload: vandien

Post on 24-Jul-2018

216 views

Category:

Documents


0 download

TRANSCRIPT

Elie  Bursztein [email protected]‐cachan.fr

A Logical Framework for Evaluating Network Resilience Against Faults and Attacks

A Logical Framework for Anticipation of NetworkA Logical Framework for Anticipation of Network Incidents

Elie Bursztein  and  Jean Goubault‐LarrecqPhd Student LSV ENS‐CACHAN CNRS INRIA DGAPhd Student  LSV ENS‐CACHAN CNRS INRIA DGA

Elie  Bursztein [email protected]‐cachan.fr

A Logical Framework for Evaluating Network Resilience Against Faults and Attacks

Outline

IntroductionNetwork EvolutionAttack Model Evolution

Anticipation game key featuresDependency relationsPlayer interactionTime 

Model LogicPositional LogicgTemporal Logic

Conclusion

Outline:  Introduction  >   Model >   Strategy >  Conclusion 2

Conclusion

28/11/2007

Elie  Bursztein [email protected]‐cachan.fr

A Logical Framework for Evaluating Network Resilience Against Faults and Attacks

The Good Old Time

Outline:  Introduction  >   Model >   Strategy >  Conclusion 328/11/2007

Elie  Bursztein [email protected]‐cachan.fr

A Logical Framework for Evaluating Network Resilience Against Faults and Attacks

The Current Internet

Outline:  Introduction  >   Model >   Strategy >  Conclusion 4

Opte project

28/11/2007

Elie  Bursztein [email protected]‐cachan.fr

A Logical Framework for Evaluating Network Resilience Against Faults and Attacks

Attack Sophistication vs. Intruder Knowledge

Outline:  Introduction  >   Model >   Strategy >  Conclusion 5

Cert/ Carnegie Mellon University

28/11/2007

Elie  Bursztein [email protected]‐cachan.fr

A Logical Framework for Evaluating Network Resilience Against Faults and Attacks

Attack Step‐stones

L t k ff lti l l bilitiLarge network may suffers multiples vulnerabilities

Patches and counter‐measures need to be prioritized 

A minor vulnerability can turn into a major holeA minor vulnerability can turn into a major hole when used as a step‐stone

Attack graph allows to reasonAttack graph allows to reason

about attack sequences

Outline:  Introduction  >   Model >   Strategy >  Conclusion 628/11/2007

Elie  Bursztein [email protected]‐cachan.fr

A Logical Framework for Evaluating Network Resilience Against Faults and Attacks

Attack Graph Example

Sandia Red Team “White Board” attack graph from DARPA CC20008 I f ti b ttl ti i t

Outline:  Introduction  >   Model >   Strategy >  Conclusion 7

Information battle space preparation experiment

28/11/2007

Elie  Bursztein [email protected]‐cachan.fr

A Logical Framework for Evaluating Network Resilience Against Faults and Attacks

Model Evolution

Anticipation

Attack graphs 

Anticipation Games

g pmodel checker(Ritchey and Ammann

2000)

Attack trees(Schneier 1999)( )

8Outline:  Introduction  >   Model >   Strategy >  Conclusion28/11/2007

Elie  Bursztein [email protected]‐cachan.fr

A Logical Framework for Evaluating Network Resilience Against Faults and Attacks

Related Work

Attack graphModel checker‐based (Ritchey et. al S&P’00, Sheyner( y , yet. al S&P’02)Graph‐based (Ammann et. al CCS’02, Ritchey et. alGraph based (Ammann et. al CCS 02, Ritchey et. al ACSAC’02, Noel et. al ACSAC’03, Wang et. al ESORICS’05, Wang et. al DBSEC’06)ESORICS 05, Wang et. al DBSEC 06)

Timed GameATL (Alur et al. 97)

The Element of Surprise in Timed Games (De Alfaro et al. CONCUR 2003)

TATL (Henzinger et al 2006 Formats)TATL (Henzinger et al 2006 Formats)

Outline:  Introduction  >   Model >   Strategy >  Conclusion 928/11/2007

Elie  Bursztein [email protected]‐cachan.fr

A Logical Framework for Evaluating Network Resilience Against Faults and Attacks

Model Key Features

• Collateral effectsD dD d Collateral effects• Trust relations

DependencyDependency

• AdministratorInteractionInteraction• Intruder

InteractionInteraction

• Action take timeTimeTime

10Outline:  Introduction  >   Model >   Strategy >  Conclusion28/11/2007

Elie  Bursztein [email protected]‐cachan.fr

A Logical Framework for Evaluating Network Resilience Against Faults and Attacks

Dependency

User Database

6

54 Email serverWeb server 54 Email serverWeb server

1 2 3

Client 3Client 2Client 1

11Outline:  Introduction  >   Model >   Strategy >  Conclusion28/11/2007

Elie  Bursztein [email protected]‐cachan.fr

A Logical Framework for Evaluating Network Resilience Against Faults and Attacks

Network interaction

Exploit vulnerabilities

Abuse trust relationsAbuse trust relations

Patch

Firewall

Restore

Outline:  Introduction  >   Model >   Strategy >  Conclusion 1228/11/2007

Elie  Bursztein [email protected]‐cachan.fr

A Logical Framework for Evaluating Network Resilience Against Faults and Attacks

Vulnerability cycle

Outline:  Introduction  >   Model >   Strategy >  Conclusion 13

Cert/ Carnegie Mellon University

28/11/2007

Elie  Bursztein [email protected]‐cachan.fr

A Logical Framework for Evaluating Network Resilience Against Faults and Attacks

Network Information

U D t b

Fixed over the time Evolve over  time

6

User Database

1 2 3 4 5 6

ρ(Public) ⊥ ⊥ ⊥ T T ⊥ρ(Public) ⊥ ⊥ ⊥ T T ⊥

ρ(Vuln) ⊥ ⊥ ⊥ T T ⊥

ρ(Compr) ⊥ ⊥ ⊥ ⊥ ⊥ ⊥

54 Email serverWeb server ρ(NeedPub) ⊥ ⊥ ⊥ T T ⊥

1 2 31 2 3

Client 3Client 2Client 1

Outline:  Introduction  >   Model >   Strategy >  Conclusion 1428/11/2007

Elie  Bursztein [email protected]‐cachan.fr

A Logical Framework for Evaluating Network Resilience Against Faults and Attacks

Information Evolution

1 2 3 4 5 6 1 2 3 4 5 61 2 3 4 5 6ρ(Public) ⊥ ⊥ ⊥ T T ⊥ρ(Vuln) ⊥ ⊥ ⊥ T T ⊥ρ(Compr) ⊥ ⊥ ⊥ ⊥ ⊥ ⊥

1 2 3 4 5 6ρ(Public) ⊥ ⊥ ⊥ T T ⊥ρ(Vuln) ⊥ ⊥ ⊥ T T ⊥ρ(Compr) ⊥ ⊥ ⊥ T ⊥ ⊥

Compr 4

ρ(Compr) ⊥ ⊥ ⊥ ⊥ ⊥ ⊥ρ(NeedPu

b)⊥ ⊥ ⊥ T T ⊥

ρ(Compr) ⊥ ⊥ ⊥ T ⊥ ⊥ρ(NeedPu

b)⊥ ⊥ ⊥ T T ⊥

28/11/2007 Outline:  Introduction  >   Model >   Strategy >  Conclusion 15

Elie  Bursztein [email protected]‐cachan.fr

A Logical Framework for Evaluating Network Resilience Against Faults and Attacks

A Incomplete Game Example

Exploit web server Patch Email serverExploit web server

E l it ilExploit email server

Patch web server Patch Email server

6

1 2 3

54

Outline:  Introduction  >   Model >   Strategy >  Conclusion 16

1 2 3

28/11/2007

Elie  Bursztein [email protected]‐cachan.fr

A Logical Framework for Evaluating Network Resilience Against Faults and Attacks

Time

Each action requires a different amount of timeEach action requires a different amount of time Patching a service: Download, extract, apply, restart

E l it iExploit a service

Firewalling a service

In anticipation games as in TATL the fastest action winwin

Player can be taken by surprise

Outline:  Introduction  >   Model >   Strategy >  Conclusion 1728/11/2007

Elie  Bursztein [email protected]‐cachan.fr

A Logical Framework for Evaluating Network Resilience Against Faults and Attacks

The element of surprise

NetworkExploit 4 in 3 unit Firewall 4 in 1 unit

Outline:  Introduction  >   Model >   Strategy >  Conclusion 1828/11/2007

Elie  Bursztein [email protected]‐cachan.fr

A Logical Framework for Evaluating Network Resilience Against Faults and Attacks

Type of attacks

Anticipation games allows to modelDenial of service

Buffer overflow execution

Permission abusePermission abuse

Cross‐scripting

Information leak

….

28/11/2007 Outline:  Introduction  >   Model >   Strategy >  Conclusion 19

Elie  Bursztein [email protected]‐cachan.fr

A Logical Framework for Evaluating Network Resilience Against Faults and Attacks

Rule Language

Outline:  Introduction  >   Model >   Strategy >  Conclusion 2028/11/2007

Elie  Bursztein [email protected]‐cachan.fr

A Logical Framework for Evaluating Network Resilience Against Faults and Attacks

Semantic

A successor node is compromisedA successor node is compromised

At least, one of the node the belongs to the equivalence is public

28/11/2007 Outline:  Introduction  >   Model >   Strategy >  Conclusion 21

Elie  Bursztein [email protected]‐cachan.fr

A Logical Framework for Evaluating Network Resilience Against Faults and Attacks

Rules example

Outline:  Introduction  >   Model >   Strategy >  Conclusion 2228/11/2007

Elie  Bursztein [email protected]‐cachan.fr

A Logical Framework for Evaluating Network Resilience Against Faults and Attacks

A Play example

User Database

6

Email serverWeb server

54

1 2 31 2 3

Client 3Client 2Client 1

Pl A ti R l T t SPl A ti R l T t SPl A ti R l T t SPlayer Action Rule Target SuccPlayer Action Rule Target SuccPlayer Action Rule Target SuccPlayer Action Rule Target SuccPl A ti R l T t SPl A ti R l T t SPlayer Action Rule Target SuccPl A ti R l T t SPl A ti R l T t SPl A ti R l T t SPl A ti R l T t SPl A ti R l T t SPl A ti R l T t SPl A ti R l T t SPl A ti R l T t SPl A ti R l T t SPlayer Action Rule Target SuccPl A ti R l T t SPl A ti R l T t SPl A ti R l T t SPlayer Action Rule Target Succ

Admin Choose Firewall 4

Intruder Choose Compromise 0day 4= 01248121620242832

Player Action Rule Target Succ

Admin Execute Firewall 4

Intruder Choose Compromise 0day 4

Player Action Rule Target Succ

Admin Choose Patch 4

Intruder Choose Compromise 0day 4

Player Action Rule Target Succ

Admin Choose Patch 4

Intruder Fail Compromise 0day 4

Player Action Rule Target Succ

Admin Choose Patch 4

Intruder Choose Compromise 0day 5

Player Action Rule Target Succ

Admin Choose Patch 4

Intruder Execute Compromise 0day 5

Player Action Rule Target Succ

Admin Choose Patch 4

Intruder Choose Compromise Forward 5 6

Player Action Rule Target Succ

Admin Execute Patch 4

Intruder Choose Compromise Forward 5 6

Player Action Rule Target Succ

Admin Choose Unfirewall 4

Intruder Choose Compromise Forward 5 65

Player Action Rule Target Succ

Admin Execute Unfirewall 4

Intruder Choose Compromise Forward 5 6

Player Action Rule Target Succ

Admin Choose Patch 5

Intruder Choose Compromise Forward 5 6

Player Action Rule Target Succ

Admin Choose Patch 5

Intruder Execute Compromise Forward 5 6

Player Action Rule Target Succ

Admin Choose Patch 5

Intruder Choose Compromise 2 5

Player Action Rule Target Succ

Admin Execute Patch 5

Intruder Choose Compromise 2 5

Player Action Rule Target Succ

Admin

Intruder Fail Compromise 2 5

Player Action Rule Target Succ

Admin

Intruder Choose Compromise 4 6

Player Action Rule Target Succ

Admin

Intruder Execute Compromise 4 6

Player Action Rule Target Succ

Admin

Intruder Choose Compromise 1 4

Player Action Rule Target Succ

Admin

Intruder Execute Compromise 1 4

Player Action Rule Target Succ

Admin

Intruder Choose Compromise 2 4

Player Action Rule Target Succ

Admin

Intruder Execute Compromise 2 4

Player Action Rule Target Succ

Admin

Intruder Choose Compromise Forward 2 5

Player Action Rule Target Succ

Admin

Intruder Execute Compromise Forward 2 5

Player Action Rule Target Succ

Admin

Intruder Choose Compromise 3 5

Player Action Rule Target Succ

Admin

Intruder Execute Compromise 3 5

Outline:  Introduction  >   Model >   Strategy >  Conclusion 23

Intruder Choose Compromise 0day 4Intruder Choose Compromise 0day 4Intruder Choose Compromise 0day 4Intruder Fail Compromise 0day 4Intruder Choose Compromise 0day 5Intruder Execute Compromise 0day 5Intruder Choose Compromise Forward 5 6Intruder Choose Compromise Forward 5 6Intruder Choose Compromise Forward 5 6Intruder Choose Compromise Forward 5 6Intruder Choose Compromise Forward 5 6Intruder Execute Compromise Forward 5 6Intruder Choose CompromiseBackward

2 5Intruder Choose CompromiseBackward

2 5Intruder Fail CompromiseBackward

2 5Intruder Choose CompromiseBackward

4 6Intruder Execute CompromiseBackward

4 6Intruder Choose CompromiseBackward

1 4Intruder Execute CompromiseBackward

1 4Intruder Choose CompromiseBackward

2 4Intruder Execute CompromiseBackward

2 4Intruder Choose Compromise Forward 2 5Intruder Execute Compromise Forward 2 5Intruder Choose CompromiseBackward

3 5Intruder Execute CompromiseBackward

3 5

28/11/2007

Elie  Bursztein [email protected]‐cachan.fr

A Logical Framework for Evaluating Network Resilience Against Faults and Attacks

Properties Semantic

Outline:  Introduction  >   Model >   Strategy >  Conclusion 2428/11/2007

Elie  Bursztein [email protected]‐cachan.fr

A Logical Framework for Evaluating Network Resilience Against Faults and Attacks

Semantic

The player A have a strategy to satisfy p y gy ythe property 

In every future the node will be compromised

28/11/2007 Outline:  Introduction  >   Model >   Strategy >  Conclusion 25

Elie  Bursztein [email protected]‐cachan.fr

A Logical Framework for Evaluating Network Resilience Against Faults and Attacks

Property Illustration

Outline:  Introduction  >   Model >   Strategy >  Conclusion 2628/11/2007

Elie  Bursztein [email protected]‐cachan.fr

A Logical Framework for Evaluating Network Resilience Against Faults and Attacks

Decidability

TATL Formula model checking in Anticipation game is decidable andAnticipation game is decidable and EXPTIME‐complete

28/11/2007 Outline:  Introduction  >   Model >   Strategy >  Conclusion 27

Elie  Bursztein [email protected]‐cachan.fr

A Logical Framework for Evaluating Network Resilience Against Faults and Attacks

One More Thing !

Elie  Bursztein [email protected]‐cachan.fr

A Logical Framework for Evaluating Network Resilience Against Faults and Attacks

It Work!

Model and Strategies are fully implemented in C

The talk example cannot be analyzed by handThe talk example cannot be analyzed by hand 4011 plays

40825 t t40825 states

28/11/2007 Outline:  Introduction  >   Model >   Strategy >  Conclusion 29

Elie  Bursztein [email protected]‐cachan.fr

A Logical Framework for Evaluating Network Resilience Against Faults and Attacks

Analyzer Demo

Elie  Bursztein [email protected]‐cachan.fr

A Logical Framework for Evaluating Network Resilience Against Faults and Attacks

Future Work

CIA model

Abstraction

Non determinism

Outline:  Introduction  >   Model >   Strategy >  Conclusion 3128/11/2007

Elie  Bursztein [email protected]‐cachan.fr

A Logical Framework for Evaluating Network Resilience Against Faults and Attacks

Conclusion

Game and Time provide a richer model for i t i l iintrusion analysis

Many directions to explorey p

Outline:  Introduction  >   Model >   Strategy >  Conclusion 3228/11/2007

Elie  Bursztein [email protected]‐cachan.fr

A Logical Framework for Evaluating Network Resilience Against Faults and Attacks

Questions

During this work  no network service was injured or tortured.  

Outline:  Introduction  >   Model >   Strategy >  Conclusion 3328/11/2007

Elie  Bursztein [email protected]‐cachan.fr

A Logical Framework for Evaluating Network Resilience Against Faults and Attacks

Rule execution time

28/11/2007 Outline:  Introduction  >   Model >   Strategy >  Conclusion 34

Rule execution time