1

GROUP CENTRIC INFORMATION SHARING USING HIERARCHICAL MODEL

By

Amit Mahale

Advisor: Dr Tim Finin

Co-Advisor: Dr Anupam Joshi

2

RISE OF INFORMATION SHARING

Need to Know v/s Need to share 9/11 commission US Federal Systems

Need to share: Uncover, respond and protect against threat

Collaborative systems examples University Environment

3

MOTIVATION

One of the central problems in information sharing is the ability to securely and differentially share information.

This issue has been addressed by Ravi Sandhu et al in their model Group Centric Information Sharing(gSIS).

Formal model for Group Centric Information sharing is available, but no practical implementation.

4

CONTRIBUTION

Develop a prototype for Group centric Information Sharing model using semantic web technologies

Modeled Hierarchical groups using OWL.

Leverage OWL’s capacity of automating group membership using Necessary and sufficient conditions

5

OUTLINE

Background : Group Centric Information Sharing

System Use-cases System Architecture System Implementation Results Algorithm Complexity Conclusion Future Work References

6

GROUP CENTRIC INFORMATION SHARING

Model developed by Ravi Sandhu et al

A first step towards a formal and systematic study of Group-Centric Secure Information Sharing Models

Brings users & objects together in a group Secure Meeting Room

7

PROPERTIES

Two types of properties

Core gSIS properties Must be enforced by all the systems modeling

gSIS.

gSIS Operations A subset of the operations may be used in the

system depending on designers discretion.

8

CORE GSIS PROPERTIES

The core properties must be satisfied by any g-SIS specificationo Persistence Properties

When a user u is authorized to access an object o, it remains the same until a group event involving u or o occurs.

o Authorization ProvenanceA user u will not be authorized to access an object o until both u and o are simultaneously group members

o Bounded AuthorizationAuthorizations do not increase during non-membership period.

G-SIS OPERATIONS

9

GROUPAuthz (u,o,r)?

Join Leave

Add Remove

Users

Objects

GROUPAuthz (u,o,r)?

Strict Join

Strict Leave

Liberal Add

Liberal Remove

LiberalJoin

LiberalLeave

StrictAdd Strict

Remove

Users

Objects

Figure courtesy Ram Krishnan et al[1]

MEMBERSHIP SEMANTICS Strict Vs Liberal Operations

User operations: <SJ, LJ> and <SL, LL> Object operations: <SA, LA> and <SR, LR>

10

SJ (u)

u not authorized to access objects added prior to join time

SA (o)

Users joining after add time not authorized to access o

LL (u)

u retains access to objects authorized at leave time

LR (o)

Users authorized to access o at remove time retain access

Figure courtesy Ram Krishnan et al[1]

11

STRICT JOIN V/S LIBERAL JOIN

During Join, If the second Join (u1; g) is an SJ.

u1 can access o4 and o5 but cannot access o2 and o3. If the Join was an LJ ,

u1 can also access o2 and o3.

During Leave SL : u1 loses access to all group objects (o1 and o2), LL: allows u1 to retain access to o2

12

STRICT ADD V/S LIBERAL ADD

During Add If (o2; g) is a SA,

Only u1 can access the object. Users u2 and u3, joining later, cannot access this object.

If (o2; g) is a LA, Current user u1 and future users u2 and u3 may access o2.

During Remove if Remove (o1; g) is an SR,

Every group user (including u1) loses access to o1. if Remove (o1; g) is an LR,

u1 can continue to access o1. However u2 and u3 will not have access to o1.

13

Operation Explaination

Strict Join(SJ) Only objects added after join time can be accessed

Liberal Join(LJ) Can access objects added before and after join time

Strict Leave(SL) Lose access to all objects on leave

Liberal Leave(LL) Retain access to objects authorized before leave time

Strict Add(SA) Only users who joined prior to add time can access

Liberal Add(LA)Users who joined before or after add time may access

Strict Remove(SR)All users lose access on remove

Liberal Remove(LR) Users who had access at remove time retain access

14

SYSTEM USE CASE

Graduate Student Admissions Promotion and Tenure Committee (P&T) Social Media Application

15

GRADUATE STUDENT ADMISSIONS

A process in which graduate student applications are scrutinized by a group of faculty members from the department.

Requirements Member should be able to access older

application. Member should not have access to documents

after leaving the groups.

16

Members join the group through ‘Liberal Join’. This will allow them to access previous

applications Applications are added with ‘Liberal Add’

Members joining the committee at a later point of time should have access to these applications.

Member leave the group using ‘Strict Leave’ Lose access to all the applications

Applications are removed from the group using ‘Liberal Remove’. Members who previously have access will still be

able to access the document.

GRADUATE STUDENT ADMISSIONS

17

PROMOTION AND TENURE COMMITTEE (P&T)

P & T committee consists of a group of full professors (tenured) who decide on the fate of an Associate professor under consideration for tenure.

Requirements Members should not have access to the P&T

documents of their senior members

18

Add the P&T documents with ‘Strict Add’ Members join the group though ‘Strict Join’/

‘Liberal Join’ If a tenured professor leaves the group, then

use ‘Strict Leave’, the documents are to be removed from the

group then use ‘Strict Remove’.

PROMOTION AND TENURE COMMITTEE (P&T)

19

SOCIAL MEDIA APPLICATION

20

SOCIAL MEDIA APPLICATION

Amit becomes a friend of Dr Finin Amit gets access to all the personal

information as well as the content (from Facebook Wall) that was shared previously

This might not be as per Dr Finin’s expectation

gSIS to the rescue

21

DR FININ, BEFORE ADDING AS A FRIEND

22

AFTER ADDING AS A FRIEND

23

WHAT GSIS CAN OFFER?

if Dr Finin adds a new friend Amit to his friend list through Strict Join: Amit will be able to access the data posted after his join time, overcoming the problem discussed in the previous slide

“Share From now” button?

Liberal Join: In addition to allowing access to new documents, Liberal

Join would allow Amit to access posts that Dr Finin shared prior to Amit’s join time through Liberal Add.

“Share Everything” button?

For Posts, Strict Add: Dr Finin should use this operation, if he wants to share the post with current set of friends and protect from his future friends.

Liberal Add: This post can be accessed by current friends as well as new friends who join at a later point of time through Liberal Add.

24

INCORPORATING GSIS INTO FACEBOOK: ADDING A FRIEND

SJLJ

25

INCORPORATING GSIS INTO FACEBOOK: ADDING A POST

Current

Current + Future

LA

SA

26

SR

LR

INCORPORATING GSIS INTO FACEBOOK: REMOVING A FRIEND

27

INCORPORATING GSIS INTO FACEBOOK: REMOVING A POST

28

COMPARISON TO CURRENT FACEBOOK MODEL

Liberal Join

Liberal Add

Strict Leave

Strict Remove

29

REVIEW

o Every user and document is associated with at least one group.

o Multiple groups may exist.o Groups may further be hierarchical.o A user may join and leave the group multiple

number of times.o A document may be added and removed from

the group multiple number of times.o The access decision of a user to a document

depends on multiple factors like Join type, Add type and the timestamps associated.

30

SYSTEM ARCHITECTURE

31

Hierarchy Ontology

Decision Engine

gSIS Rules

Inferred Data

Group data

Results

gSIS Ontolog

y

Access decisions

Reasoning

Reasoning

SYSTEM ARCHITECTURE

32

GROUP OPERATION DATA

Data about the group members/documents and their operations.

Group user can join and leave the group multiple numbers of times

<user_id>,<join_time>,<join_type>,<leave_time>,<leave_type>, <group_name>

<doc_id>,<Add_time>,<Add_type>,<Remove_time>,<Remove_type>, <group_name>

33

HIERARCHY ONTOLOGY

Used to represent the hierarchy of the system

Helps to infer the additional groups that the member belongs to

In a hierarchy of Professor, Asst Professor and Lab Instructor.

An user added to a Professor group should by default have access to the documents added to Asst Professor and Lab Instructor group.

34

CEO

CTO

Project Manager

Team Lead

Associate Engineer

CFO

Finances Team

HIERARCHY IN GROUPS

Disaster Management GroupFire

Fighters

Police Department

Ambulance

35

MOTIVATION FOR USING SEMANTIC WEB

System Understandable

Usage of Ontology makes the system flexible and extendable.

gSIS is modeled using temporal logic, thus developing the prototype using OWL(based on logic) helps to prove the correctness of the model.

36

INFERRED DATA

The RDFS reasoner is used to infer additional groups to which the user belongs to; using the hierarchy ontology.

The inferred data along with the Group data is then fed to the decision engine.

37

GSIS ONTOLOGY

38

DECISION ENGINE

Central system of the gSIS model

Every access decision depends on the combination of group operations and the timestamp’s associated with them.

The rules are modeled to cover all combinations of events that can occur in a group centric information sharing environment.

39

STRICT JOIN, STRICT ADD, STRICT LEAVE, STRICT REMOVE

Let Uj & UL be the User Join and Leave time and

DA & DR be the Document Add and Remove time

User Join (Uj)

Doc Add (DA)

Access time[DA – Min (UL, DR)]

User Leave (UL)

Doc Remove (DR)

40

LIBERAL JOIN, LIBERAL ADD, LIBERAL LEAVE, LIBERAL REMOVE

Let Uj & UL be the User Join and Leave time and

DA & DR be the Document Add and Remove time

User Join (Uj)

Doc Add (DA)

Access time[Max(UJ,DA) – Max (UL, DR)]

User Leave (UL)

Doc Remove (DR)

41

STRICT JOIN, LIBERAL ADD, STRICT LEAVE, LIBERAL REMOVE

Let Uj & UL be the User Join and Leave time and

DA & DR be the Document Add and Remove time

User Join (Uj)

Doc Add (DA)

Access time[DA –UL]

User Leave (UL)

Doc Remove (DR)

42

LIBERAL JOIN, STRICT ADD, LIBERAL LEAVE, STRICT REMOVE

Let Uj & UL be the User Join and Leave time and

DA & DR be the Document Add and Remove time

User Join (Uj)

Doc Add (DA)

Access time[DA –DR]

User Leave (UL)

Doc Remove (DR)

43

CONCLUDE DECISION ENGINE

Can observe a pattern Check for conformance with gSIS operations properties Compute access start time Compute access end time.

Constructing the rule becomes tedious and complex to handle in OWL. Our prototype uses an pragmatic approach, Semantic web + procedural method.

Semantic Web technology to represent and reason about the hierarchy; Procedural method to compute access decisions relying on the gSIS semantics.

44

AUTOMATING GROUP MEMBERSHIP

Automatically classifies users to relevant groups.

Leverages OWL feature of Necessary and Sufficient conditions.

Whenever a user satisfies the N&C, the user is added to the group.

45

EXAMPLE

A Professor is added to the UMBC CS Tenure committee if He/She is a Full Professor A Professor @ UMBC. Faculty in the CS DepartmentThe ontology is as follows

46

N & C

N & C

N & C

AUTOMATING GROUP MEMBERSHIP

47

AUTOMATED DOCUMENT CLASSIFICATION

Documents are classified as Top Secret, Secret, Confidential, Restricted, Unclassified.

Groups can be governed by policies on the type of documents added to each group.

Utilizes OWL Features and Hierarchy resolution

48

‘War room’ group contains all documents from level ‘ Top Secret’ and below.

‘Air Force’ group ‘Top Secret’ ‘ Air Force’ domain.

‘Air Force Research’ group ‘Air Force’ domain Unclassified

49

SYSTEM IMPLEMENTATION

50

SYSTEM IMPLEMENTATION

53

RESULTS

54

VALIDATION

We develop sample data set for the P & T use case

o To demonstrate hierarchical groups, we have two groups, ‘Tenure group’ and ‘Associate Professor Group’

o Data contains details about members and their documents.

o Rule : Tenure group members have access to the documents of ‘Associate Professor group’

55

QUERIES

56

QUERY 1: USER-DOCUMENT-TIME

Did Dr Finin have access to Dr Joshi’s Tenure file in 2005?

Access Granted

57

QUERY 2: USER ACCESS DETAILS

List all the documents that Dr Finin has access to

58

QUERY 3: DOCUMENT ACCESS

List all the users who have access to ‘Andrewdoc'[Andrew is an Assistant Prof and under consideration for tenure]

59

QUERY 4: TIME BASED ACCESS

List all the documents that were accessible to users in 1994

60

QUERY 5: USER-DOCUMENT

Did Dr Finin ever have access to Nicholasdoc?

61

ALGORITHMIC COMPLEXITY

n users m documents Computing Access intervals would take n*m O(nm) when m=n O(n2)

Whenever group membership changes User joins the group: (1 * m) O(m) Document is added to the group: (n * 1) O(n)

62

CONCLUSION

We have presented a agile framework for secure information sharing.

We have also modeled gSIS to support hierarchical groups and opened up opportunities to extend gSIS in several dimensions like automated group membership.

Finally we have demonstrated the usefulness of gSIS in real world applications.

63

FUTURE WORK

Develop the administrative model for gSIS.

Write policies to enforce the gSIS operation semantics.

64

65

66

REFERENCES[1]Ram Krishnan, Ravi Sandhu, Jianwei Niu and William Winsborough, Foundations for Group-Centric Secure Information Sharing Models. Proc. 14th ACM Symposium on Access Control Models and Technologies (SACMAT), Stresa, Italy, June 3-5, 2009, pages 115-124. [2] Ram Krishnan, Ravi Sandhu, Jianwei Niu and William Winsborough, Towards a Framework for Group-Centric Secure Collaboration. In Proc. 5th IEEE International Conference on Collaborative Computing: Networking, Applications and Worksharing (CollaborateCom), Crystal City, Virginia, November 11-14, 2009, pages 1-10.   [3] Ravi Sandhu, Ram Krishnan, Jianwei Niu and William Winsborough, Group-Centric Models for Secure and Agile Information Sharing. In Proceedings 5th International Conference, on Mathematical Methods, Models, and Architectures for Computer Network Security, MMM-ACNS 2010, St. Petersburg, Russia, September 8-10, 2010, pages 55-69. Published as Springer Lecture Notes in Computer Science Vol. 6258, Computer Network Security (Igor Kotenko and Victor Skormin, editors), 2010. [4] T. Finin, A. Joshi, L. Kagal, J. Niu, R. Sandhu, W. Winsborough, and B. Thuraisingham, ROWLBAC - Representing Role Based Access Control in OWL, Proceedings of the 13th

ACM symposium on Access Control Models and Technologies, ACM Press New York, June 2008.

67

[5] Anne Cregan, Malgorzata Mochol, Denny Vrandecic, Sean Bechhofer Pushing the limits of OWL, Rules and

Protégé. A simple example Workshop - OWL: Experiences and Directions (OWLED-2005), Galway, Ireland,

November 2005 [6] R. Sandhu et al, Role-Based Access Control Models, IEEE Computer, 29(2):38-47,Feb 1996, 

Google Scholar Search [7] R. Sandhu and P. Samarati, Access Control: Principles and Practice, IEEE Communications, 32(9): 40-48, Sept. 1994, Google Scholar Search

[8] Semantic web: http://www.w3.org/2001/sw/ [9] Bechhofer, S.; van Harmelen, F.; Hendler, J.; Horrocks, I.; McGuinness, D.; Patel-Schneider, P.; and Stein, L. 2004. Owl web ontology language reference. w3crecommendation. [10] United States Intelligence community ‘INFORMATION SHARING STRATEGY’, OfficeOf the Director of National Intelligence, http://www.dni.gov/reports/IC_Information_Sharing_Strategy.pdf [11] Jones, H., and Soltren, J. 2005. Facebook: Threats to privacy. 

REFERENCES