SPIG

Security by Privileges, Inspection and Guaranty for Ambient Agents Group

Nardjes Bouchemal

1,2

and Ramdane Maamri

LIRE Laboratory, Computer Science Department, Mentouri University of Constantine, Constantine, Algeria

University Centre of Mila, Mila, Algeria

Keywords: Ambient Intelligence, Ambient Agents Group, Security, Inspection, Guaranty.

Abstract: In a future vision of Ambient Intelligence – or AmI – our surrounding environment will integrate a

pervasive and interconnected network of devices equipped with sensors, actuators and ambient agents.

However, it is a big security issue when building this kind of complex systems, and it is not enough that

ambient agent platform provides a set of standard security mechanisms such as sandboxing, encryption and

digital signatures. Furthermore, we must take into account ambient agent limitations and we mustn’t endow

it with complex cryptography concepts or historic data. This is why our approach, proposed in this paper, is

based on cooperation between group members and behavior inspection of new integrated agents, in addition

of some cryptography concepts. The proposed protocol is based on cooperation, collective decision,

guarantee and attribution of privileges according to the trust degree of agents.

1 INTRODUCTION

The first vision about pervasive computing was

proposed by Mark Weiser in his famous article “The

Computer for the 21st Century” 1991 (Weiser,

1991). His vision of computers fully integrated in

the human environment and gracefully providing

information and services to users is still an open

issue in computer science and computer engineering.

The ambient intelligence paradigm builds upon

pervasive computing, ubiquitous computing,

profiling practices, context awareness, and human-

centric computer interaction (Augusto, 2009),

(Weber, 2005).

Consequently, the development of AmI

environment needs robust technologies, such as

Agents (Ferber, 1995), that respond to AmI

characteristics.

They have already been successfully applied to

several AmI scenarios such as education, culture,

entertainment, medical domain, robotics and home

(Tapia, 2010). In these systems, agents are called

ambient agents.

However, developing agents in complex AmI

environments leverages the common issues of

distributed applications. It also poses new security

challenges to handle the dynamicity of such

environments. Indeed, ambient agents will operate

on portable electronic devices with little autonomy

energy, limited computing and storage capabilities

that vary from one device to another (Ko and

Ramos, 2009).

Thus, ambient agents groups are at risk if there is

no control of different participants, or agents who

require communicating. The crucial question is how

to ensure any confidentiality, integrity and privacy

of exchanged messages and data if, at the outset, we

are not sure that we communicate with the correct

entity? (Stajano, 2010)

This article presents an approach to protect

ambient agents groups by

taking in to consideration

the new challenges required by ambient intelligence

environment.

In section two we sketch out security

approaches for classic multi agent systems and we

discuss their insufficiency to AmI field. In section

three we define the protocol, based on guaranty,

inspection of new integrated agents, on cryptography

concepts and attribution of privileges according

the trust degree of integrated agent. Section four

presents implementation issues using JADE platform

and summarizes preliminary comparative results.

Finally, a conclusion recaps the paper and future

works.

318

Bouchemal N. and Maamri R..

SPIG - Security by Privileges, Inspection and Guaranty for Ambient Agents Group.

DOI: 10.5220/0004261703180323

In Proceedings of the 5th International Conference on Agents and Artiﬁcial Intelligence (ICAART-2013), pages 318-323

ISBN: 978-989-8565-39-6

 2013 SCITEPRESS (Science and Technology Publications, Lda.)

2 RELATED WORK

2.1 Multi Agents System Protection

The goal of this section is to present several viable

technologies that can be used in multi-agent systems

in order to provide a secure infrastructure.

Then we discuss the robustness of these

approaches to respond to different requirements of

ambient intelligence.

The Protected Computing approach is based on

the partitioning of the software elements into two or

more parts. Some of these parts (called private parts)

are executed in a secure processor, while others

(public parts) are executed in any processor even if it

is not trusted.

In general, the Protected Computing model requires

the use of secure coprocessors that have asymmetric

cryptography capabilities; secure storage to contain

a key pair generated inside the coprocessor and

ensure that the private key never leaves it (

Maña

2006).

Trusted Computing Platforms (TCG, 2005) take

the advantage of the use of a hardware element in

order to provide a secure environment. These

hardware elements are called TPM (Trusted

Platform Modules). A TPM is a microprocessor with

some special security features. The key idea behind

the Trusted Computing Platform is to build a trusted

environment starting from the TPM and extending

the trust to the rest of the system elements. At the

beginning of the execution the only trusted element

in the system is the TPM.

The technique called proof-carrying code (PCC)

(Necula, 1997) is a general mechanism for verifying

that a software element can be executed in a secure

way. For this purpose, every code fragment includes

a detailed proof called code certificate.

Maña

et al. in (Maña 2006), proposed an approach to

protect a society of collaborating agents, by making

every agent collaborate with one or more remote

agents running in different hosts. These agents act as

secure coprocessors for the first one. Likewise, these

agents are in turn protected by other agents.

The Static Mutual Protection strategy can be

successfully applied to many different scenarios.

However, there will be scenarios where (i) it is not

possible to foresee the possible interactions between

the agents at development time, (ii) where the agents

are generated by different parts.

Maña

et al. in (Maña 2004) and (Maña 2007)

proposed a new strategy called Dynamic Mutual

Protection where each agent is able to execute

arbitrary code sections on behalf of other agents in

the society. Each agent includes a public part, an

encrypted private part and a specific virtual

machine.

2.2 Discussion

Ambient agents have limited capabilities due to their

closely attachment with the device capabilities. They

are cognitive, but hold little knowledge, don’t

require much resources and form very simple plans

especially when they are embedded in very small

devices like sensors.

Furthermore, in the context of ambient intelligence,

entities are intelligent and can pretend being

confident just to be integrated to the ambient agents

group. Then, they can make dangerous actions and

may even lead to the destruction of the group.

Consequently, the question to ask is: can

presented approaches respond to ambient agents

properties?

Previous presented works don’t inspect agents after

integration; they just verify the origin and the

authenticity. But even the origin is trustworthy;

agent can be altered during its displacements.

Furthermore, presented approaches use complex

cryptographic concepts, which make them robust,

but more difficult to realize in case of limited

ambient agents’ resources. Moreover, these

approaches are limited by the number of agents

included in the group, where the management

becomes more and more difficult.

In conclusion, during the design of ambient

agent’s security, we must take in mind ambient

agents’ limitations and we mustn’t endow them with

complex cryptographic concepts or historic data.

This is why our approach is based on cooperation,

collective decision and division of security tasks

between agents of the group.

3 SPIG: SECURITY BY

PRIVILEGES, INSPECTION

AND GUARANTY

We proposed in previous works (Bouchemal, 2012)

an approach based on guaranty, where an agent is

integrated into a group only if an agent from the

group knows it. The problem is posed when no agent

knows the new one. In this section, we bring

ameliorations and consider the case where an agent

is completely unknown.

SPIG-SecuritybyPrivileges,InspectionandGuarantyforAmbientAgentsGroup

319

3.1 Inspiration

In the design of the protocol, inspiration was taken

from the human behaviour and thinking. In order to

more understand the idea, we present in this section

a real life scenario. Let’s consider a home where

lives a couple and their children: Charlie and Betty,

having consecutively fifteen and thirteen years old.

The father works in a computer company with other

persons chaired by a director. On the other hand,

Charlie and his sister attend a school, near their

house, managed by a principal. They have several

friends and teachers within the school.

One day, the door rings, the father opens. He

finds a teenager claiming to be Charlie’s friend. The

father calls his son and asks him.

Charlie confirms that he knows the visitor. The

father brings Charlie’s friend, but he is suspicious,

so he restricts visitor displacements in the house and

recommends to Betty and her mother to keep an eye.

Another day, the father invites some colleagues

to have lunch in the home. After that, he takes them

to visit the garden and to drink coffee there. They

are supervised just by the father (eventually by the

mother) because these gents are supposed

trustworthy.

3.2 Departure Points

We consider a group of ambient agents embedded in

mobile devices and represent an object and/or a user.

We define some departure points in what follows:

 Each member of the group has a set of

characteristics:

 Id_Agent: A unique identification within

the group.

 Priv: Privileges within the group (set of

permitted actions).

 LiD: The list of identifications agents of the

group.

 TrustDeg: A degree of trust within the

group: doubtful, malicious or trust.

 ResidTime: The estimated residence time

within the group.

 DevRep: Representation of the device

resources: energy, memory and processor

capabilities.



Each group has a specific and trust agent called

representative,

having maximum privileges,

powerful device and with longer time of

residence.



Representative of the group has the instantly list

of all group agents with their priviliges, trust

degrees and behavior reports.



Members of the same group know each other

and cooperate to complete various tasks. They

share a Common Public Key (CPK) known only

by group members and frequently changed by

the representative.



Members have communication keys (Cki for an

agent i) to establish secure channels with the

representative. Cki is frequently changed.



The group have two vice representatives having

most privileges, residence time and powerful

device. The goal is to ensure the suitable

conduct of the group in case of attacks against

the representative.



An agent can belong to two groups at once, or

more (Figure.1). In this case it knows agents and

have keys of both groups.

Figure 1: Agents belonging to more than one group.

3.3 The Protocol

The key idea is based on behaviour inspection of

new integrated agents, cryptography concepts,

guarantee and attribution of privileges.

When an external agent requests communication

or data from the group, it contacts either a member

or the representative. If it contacts a member, this

one redirects the request to the representative. If the

external agent contacts directly the representative, it

cooperates with the group’s agents to verification,

integration or rejection of the request, and inspection

after integration.

3.3.1 Verification

When the representative group receives an

integration demand from an external agent, it sends

its identity to group members. We study two cases

before responding: First, external agent is known by

at least one agent from the group (so called The

guarantor), when both agent belong to another

group (Group 2 in Figure.1).

Group1

Group2

Group3

Rep1

Rep2

Rep3

Guarantor

External Agent

ICAART2013-InternationalConferenceonAgentsandArtificialIntelligence

320

The guarantor sends a confirmation to

representative who will add the external agent to the

group. The integration is done differently depending

on the privileges of the guarantor.

In the second case, external agent is unidentified

by any agent of the group. Representative delegates

a committee to examine it. The committee is a

subset of agents with most privileges, resources,

residence time and trust degree.

Committee members use sandbox technique

(Gong, 1997) by sending different trivial code and

data to the external agent. After execution, it resends

results, code and data. Agents of the committee

check the code, data and results and send their

reports including attributed trust degree and

privileges to representative who makes the last

decision. If the trust degree is malicious, the request

is rejected. Else, if the degree is doubtful, the request

is accepted but with minimum priviliges and the

agent is inspected. If the degree is trust, the request

is accepted and the agent is integrated, but must also

be inspected.

3.3.2 Integration

The new agent integration within the group is done

only if trust degree is not malicious. First,

representative establishes a secure channel with the

new agent by sending a communication key Ck’.

Then it sends the common public key: Ck’(CPK),

and the identities list of group members (LiD). It

also sends new agent identity, trust degree and

privileges to members. After that, the representative

informs all members that the integrated agent must

be inspected.

3.3.3 Inspection after Integration

Members cooperate to keep track of new integrated

agents in order to confirm their intentions. They

oversee the behavior of integrated agent and they

make a report to the representative, if they discover

malicious or doubtful behavior. The representative

can make changes in status, and privileges

(Figure.2).

Figure 2: Possible changes of integrated Agent’s status.

It can revoke agents having malicious behavior by

changing CPK and adding it in its black liste. After

any changes, the representative informs all group

members.

4 IMPLEMENTATION AND

EXPERIMENTATION

To validate the proposed protocol, we conduct, in

section 4.1, an implementation using JADE platform

(Java Agent DEvelopment framework) (Bellifemine,

2004), (Caire, 2007). Jade is Java-based and FIPA

(Foundation for Intelligent Physical Agents)

compliant development framework. In section 4.2,

we establish a comparison between our protocol and

the Protection Computing Approach.

4.1 Implementation of Groups and

Agents

Agents have been designed so that they are simple,

flexible, and so that an agent with the same structure

can run both on a simple processor assigned to a

sensor and on a powerful computer. The agents are

cognitive, but hold little knowledge and form very

simple plans.

Figure 3:Agents and Groups.

We would like to simulate the scenario presented in

section 3.1. Thus, we create three containers each

one represents a group: main container for group1,

modelling the father’s work group and comprises

AgentDirector (the representative of the group),

AgentFather, Agent1G1, Agent2G1 and Agent3G1

(Figures.3 and Figure4). Container1 represents

Charlie’s home group and includes AgentFather (the

representative of the group), AgentMother,

AgentChild1 (for Charlie) and AgentChild2 for his

sister. Finally, container3 represents Charlie’s school

group, icluding: AgentScholPrincipal (the

representative), AgentChild1, AgentStudent1,

AgentStudent2 and AgentStudent3.

Agents are embedded in various devices (laptops

SPIG-SecuritybyPrivileges,InspectionandGuarantyforAmbientAgentsGroup

321

for the representatives, PDA and smart phones for

the rest of agents).

Figure 4: Agents and containers using JADE.

To implement agent’s behaviours in JADE, we

use behaviour composition by building simple

behaviours. The basis for this feature is provided by

the CompositeBehaviour class included in the

jade.core.behaviours package.

Messages exchanged by agents are instances of the

jade.lang.acl from ACLMessage class, which

represents ACL messages. It contains a set of

attributes as defined by the FIPA specifications.

4.2 Experimental Scenes

To verify the efficiency of the proposed protocol, we

advise experimental scenes using SPIG protocol and

the Protected Computing approach.

As mentioned in section 2.1, the Protected

Computing Approach uses a trusted processor to

enforce the correct execution of the private parts of

the program.

Therefore, using the protected computing model,

the code of each agent is divided into public and

private parts. For reasons of simplicity, we will

consider that the code of each agent is divided in

two parts: a public one and a protected one.

4.2.1 Scene One: The External Agent is

Known

AgentStudent1 from Container2 requests integration

into Container1; it contacts AgentFather (the

representative). AgentChild1 knows AgentStudent1,

so it can be integrated with same AgentChild1’s

privileges, and must be inspected by all members:

AgentFather, AgentMother and AgentChild2.

4.2.2 Scene Two: Malicious Action from

AgentStudent1

After integration, AgentStudent1 becomes untrusted.

We simulate this behavior by injecting a malicious

code into AgentMother, which discovers this action

and contacts AgentFather to revoke it.

4.3 Experimental Results

In scene one, we have calculated integration time of

external agent, when it is known, relation to the

number of agents in the group. We have established

comparison graph, as shown in Figure.5.

We can observe that the latency taken by the

cryptographic operations, division of agent’s code

and the communication with the coprocessor, is an

important factor that affects the performance of the

protected computing approach.

In SPIG protocol, integration time is not so

considerable, because when the agent is known, it is

integrated faster even the number of agents is big.

Figure 5: Comparison of agent’s integration time.

Concerning scene two, we calculate necessary

time to detect AgentStudent1 attack using SPIG

protocol then the protected computing approach

relation to increasing number of integrated agents.

For protected computing approach, when the

number of agents increases, the management

becomes more difficult, principally because of

devices resources’ limitation, rising of divided parts

of agents and the overflow of trusted processor

(Figure.6). These circumstances make the

0,5

1,5

2,5

3,5

5 10152025

IntegrationTimeins

NumberofAgents

ProtectedComputingApproach

SPIGProtocol

ICAART2013-InternationalConferenceonAgentsandArtificialIntelligence

322

verification of malicious actions slower.

Figure 6: Comparison of attack detection time.

5 CONCLUSIONS

We presented in this paper an approach to protect

ambient agents group, where each agent is

embedded to a device and represents a user.

If a new agent wants to communicate and share

data with agents of this group, it must be verified,

integrated if it is not malicious, and inspected after

integration. These functions are done cooperatively

by the members.

We have detailed the protocol and introduced a

set of encryption keys: CPK (Common Public Key)

shared and known by all members of a group, Cki

(Communication Key for Agent i) sent by the

representative to an agent i. We presented a simple

case study in health domain, and finally we

presented some implementation issues using Jade

platform. Our on-going work is to apply the

proposed approach in a real case and real devices, by

using a professional AmI platform. We would like to

test the reliability in large scale.

REFERENCES

Augusto, J. C., 2008. Ambient Intelligence: Basic

Concepts and Applications. In Computer and

Information Science: Software and Data

Technologies, Volume 10, Part 1. Berlin, Germany:

Springer Berlin Heidelberg.

Bellifemine, F., Caire, G., Trucco, T., Rimassa, G., 2004

Jade Programmer’s Guide, version 3.2 . The book.

Bouchemal, N., Maamri, R., 2012. SGP: Security by

Guaranty Protocol, DCAI 12, AISC 151, pp. 289-296,

Caire, G., 2007. JADE tutorial: JADE programming for

beginners, Telecom Italia Laboratory.

Ferber, J., 1995. Les Systèmes Multi-Agents.

Informatique Intelligence Articielle, The book,

InterEdition.

Gong, L., Mueller, M., Prafullchandra, H., Schemers, R.,

1997. Going Beyond the Sandbox: An Overview of

the New Security Architecture in the Java

Development Kit 1.2, In Proceedings of the USENIX,

California.

Ko, H., Ramos, C., 2009. A Study on Security Framework

for Ambient Intelligence Environment (ISyRAmISF :

ISyRAmI Security Framework)", in Fifth

International Conference on Wireless and Mobile

Communications.

Maña, A., López, J., Ortega, J., Pimentel, E., 2004. A

Framework for Secure Execution of Software.

International Journal of Information Security.

Maña, A., Muñoz, A., 2006. Mutual Protection for

Multiagent Systems. In: proceedings of the Third

International 3rd International Workshop on Safety

and Security in Multiagent Systems (SASEMAS ’06)

Maña, A., Muñoz, A., Serrano, A., 2007. Towards Secure

Agent Computing for Ubiquitous Computing and

Ambient Intelligence. UIC 2007, LNCS 4611, pp.

Springer-Verlag Berlin Heidelberg.

Necula, G., 1997. Proof-Carrying Code. In: Proceedings

of 24th Annual Symposium on Principles of

Programming Languages.

Stajano, F., 2010. Security Issues in Ubiquitous

Computing, Handbook of Ambient Intelligence and

Smart Environments, Springer LLC 2010.

Tapia, D., 2010. Agents and ambient intelligence: case

studies. Journal of Ambient, Intel Human Comput.

TCG, 2005. Trusted Computing Group: TCG

Specifications. Available online at,

https://www.trustedcomputinggroup.org/specs/

Weber, W., Rabaey, JM., Aarts, E., 2005. Ambient

Intelligence. Springer, New York.

Weiser, M., 1991. The Computer for the 21st Century,

Scientific American, 265(3):94–104,

5 10152025

DetectioTimeins

NumberofAgents

ProtectedComputingApproach

SPIGProtocol

SPIG-SecuritybyPrivileges,InspectionandGuarantyforAmbientAgentsGroup

323