Security in Mobile Agents: Advantages, Applications, and Threats, Exercises of Mobile Computing

Download Security in Mobile Agents: Advantages, Applications, and Threats and more Exercises Mobile Computing in PDF only on Docsity!

Security in Mobile

Agents

Presented

By

Padma Kapse

OUTLINE

Introduction

Advantages of Mobile Agent paradigm

Applications of Agents

Security Issues

Classification of Security threats

Security Measures

Examples of Mobile Agent Systems

Simple Model of an Agent

Agent Platform

Agent Platform

Agent Platform

Agent

Agent

Network

Agent

Agent to Agent

communication

Advantages of an Agent System

Reduces network load and latency – There is usually no transmission of

intermediate result. This conserves the network bandwidth.

Asynchronous – Since the agents are autonomous, the mobile device that

dispatches the agent need not be connected all the time.

Fault Tolerant – If one of the host is down, the agent can be transferred to

another host for execution. The agents can be programmed to adapt

dynamically to the network conditions.

Can be customized according to the needs.

Can be deployed in a heterogeneous environment, as only the execution

environment is of concern not the specifics of the Host Platform.

Parallel processing – agents can execute concurrently in a distributed

system. A single task can be decomposed amongst multiple agents.

Network Security testing - Agents can be used in network intrusion

detection.

Database searches – It is efficient to send out an agent to perform a

specialized search on a large database than to move big chunks of data to the

client using up enormous amount of bandwidth.

Security Issues in Agent Systems

The autonomous and mobile nature of the agents introduces new complexities

and security issues such as

In E-Commerce scenario, Agent may be carrying sensitive information like

Social Security Number or Bank Account details. Agents must be secure and

tamper-proof, and must not reveal information inappropriately.

The Host platform should provide a safe environment for the agent to execute.

A malicious agent may attack the Host and access sensitive data or tie up

inordinate amount of resources causing Denial of Service to other applications

or agents.

Since agents traverse multiple hosts trusted to different extents, implementing

any security measure is complicated.

Security Measures

Security in Agent System is based on the principle of trust. A set of security

policies and protocols establish the trust relationship between the entities.

It is assumed that the agent trusts the Home platform that dispatches it.

Agent attacking the Host environment

Traditional methods such as authentication, access control, sand-boxing

techniques, cryptography can be used to secure the Host.

Authentication and access control mechanisms – This is the first line of

defense against a malicious agent. If the Host can authenticate the agent and in

turn the device that dispatched the agent, it can apply authorization and access

control.

Safe Code Interpretation – Due to the necessity for the agents to run on

heterogeneous computer , interpreted scripting or programming languages are

used. This produces intermediate code that is executed by a virtual machine

that sits on top of the native processor and OS. This virtual machine can

enforce additional security.

Path Histories - An agent could reach the host by making a number of

hops. During this transit a malicious host could have morphed the agent into

a malicious agent. By storing the log of the travel of the agent, the current

host can determine the route taken by the agent.. Each host platform to

which the agent travels to, appends a signed entry to the path. This entry

indicates the hosts identity as well as the identity of the next host the agent

intends to visit. The platform has to judge by looking at the log if the

previous platforms can be trusted.

State Appraisal – The author of the agent supplies a state appraisal function

called maximum function. This function calculates, depending on the state

of the agent, the maximum set of permissions to be granted to the agent.

This function is packaged together with the agent. The user/owner of the

agent also supplies another state appraisal function called the request

function. This calculates the permissions the user wants the agent to have

during execution. The host platform uses these state functions to verify the

correct state of the agent and hence determines the privileges to give to the

agent depending on its state. This ensures that the agent has not turned

malicious due to alterations of its states.

Obfuscated code – A “blackbox” agent is generated from the agent

specification wherein the agent’s code and data cannot be read or modified.

Only its input and output can be observed. The algorithm that creates the agent

is called “mess-up or obfuscating algorithm”. To prevent dictionary attacks the

algorithm, that converts the agent specs into an agent, uses some random

parameters. These parameters allow creation of number of different agents out

of the same specification. The agents differ in code and data representation but

give the same results.

Secure Routing - An agent can be programmed to have a routing policy such

that it migrates only to certain servers. Since a malicious host can tamper with

the agent’s itinerary and also computation results, which can propagate, some

fault tolerance is needed to ensure that the agent reaches its destination and

perform its job correctly. Replication and voting can be used to achieve fault

tolerance. The agent is replicated at each stage and run on hosts. The results

from these computations are compared (i.e. voted). Then the correct result is

sent out as output.

Detecting attack using Dummy data –In this technique, dummy data items

called detection objects are used. This dummy data is stored in the database of

the agent and it will not be modified while the agent performs its functions.

After the agents return, if the detection objects have not been modified, then

one can have reasonable confidence that legitimate data also has not been

corrupted. This technique requires that the dummy data should not adversely

affect the results of the query.

Using Trusted Hardware - This technique uses tamper proof trusted

hardware to encapsulate the entire agent execution environment in which the

agent executes, thus isolating the agent from the malicious host. The whole

agent is not visible to the host environment. The agent in this system will

interact with the Host environment through messages. Each Host in the Mobile

Agent System is equipped with this hardware. The hardware can be in form of

PC Cards, Smartcards, Integrated Circuits, etc. PC Cards are powerful and

allow the whole agent code to be loaded into the card. Smartcards are limited

in their capabilities. Only a part of agent code can be loaded on the card. The

agent carries rest of the code along with it. The code that the agent carries is

encrypted.

References

[Che] Mobile Agents: Are they a good idea? , David Chess, Colin Harrison, Aaron

Kershenbaum: http://citeseer.nj.nec.com/chess95mobile.html

[

Funf] Protecting Mobile Web-Commerce Agents with Smartcards , Stefan Funfroken,

Dept of CS, Darmstadt Univ of Tech. , Germany :

http://citeseer.nj.nec.com/unfrocken99protecting.html

[Hohl] Time Limited Blackbox Security:Protecting Mobile Agents From Malicious

Hosts, Fritz Hohl:

http://citeseer.nj.nec.com/hohl98time.html

[Kar] Network Security Testing Using Mobile Agents, T. Karygiannis,

[email protected], NIST:

http://csrc.nist.gov/mobilesecurity/Publications/Agents_PAAM98.pdf

[Mea] Detecting Attacks on Mobile Agents, Catherine Meadows, Center for High

Assurance Computing Systems, Washington DC:

http://citeseer.ist.psu.edu/meadows97detecting.html

[Mins] Cryptographic support for Fault-Tolerant Distributed Computing, Yaron

Minsky et al : http://citeseer.nj.nec.com/minsky96cryptographic.html

[Ous] The Safe-Tcl Security Model, J.K.Ousterhout et al, Tech. Report SMLI TR-

97-60, Sun Microsystems,1997: h

ttp://citeseer.nj.nec.com/ousterhout97safetcl.html

[Sta]Is it an Agent, or just a Program? : A Taxonomy for Autonomous Agents,

Stan Franklin and Art Graesser , Institute for Intelligent Systems,

University of Memphis :

http://www.cs.memphis.edu/~franklin/AgentProg.html

[Tom] Protecting Mobile Agents against Malicious Hosts, Tomas Sanders and

Christian Tschudin. Proceedings of Workshop on Mobile Agents and Security,

number 1419 in LCNS, pages 44-60, 1997:

http://citeseer.nj.nec.com/sander98protecting.html