Well thatís agents in a single chapter. Not
an easy thing to do. So what benefit will agents bring, and are
they worth the trouble? Well theyíre worth it as they allow us to
migrate processing away from busy servers towards the client. The
allow tend to carry out their tasks quietly and can turn raw data
into a form that the server can quickly use. But, arenít we leaving
ourselves open to a new wave of super viruses, in the form of undercover
agents and undercover servers. These little agent programs work
for the other side (the hacker) and can pass on sensitive information
to others. The Internet now allows for a convenient path for untrusted
agents to travel. So how can we stop this, well the only real way
of to agents to authenticate themselves to the server, and vice
versa. This will involve some form data encryption, possible using
a secret key. Agents cannot also be allows to roam wherever
they want, thus there must be some mechanism for providing homes
for agents to live, which only allows valid agents to live in the
big worry is the use of agents to breach civil liberties. Whatís
to stop a government agency from downloading an agent to your home
computer, which then monitors every event within the computer, and
finds out the contents of all of your emails? It possible, and there
are many commercial programs which will scan a computer looking
for available ports to connect into. Once attached to the computer
there is little to stop a downloaded program from gain access to
all the resources of the computer. At present Java programs which
run from WWW pages are protected against this type of attack, and
only allow minimum access to local resources.
beware, the cleanest attack on a system is through the TCP/IP stack.
If these is tampered with it can allow for programs to be run which
open up local ports which can be connected into whenever the user
logs into the Internet (or even any network).
what the next logical step in client-server networks: agent technology,
and whatís the next natural step in agent technology: mobile agents.
These helpful little agents like to work independently. They are
dispatched to clients, and then work quietly gathering information,
and sending it back whenever required, or whenever a user connects
back onto the Internet. They are thus extremely useful when users
are also mobile, and use notebook computers to perform their business.
what about security? Passwords and login IDs are a terrible method
of securing a system. They provide little protection against external
hackers. An improved method for a server to scan an audit log file
for the user and determine their typical usage (their user profile).
An agent can then be dispatched to the computer which the user is
using, which then checks to see if the user is operating as they
usually do. If they do not, the agent can alert the server that
there is a possible breach, or that the user may be acting in an
usual manner (typically the first signs of a fraud). An example
would be if a user started to type at 70wpm where before they used
chop-sticks to type their commands. Agents can also be dispatched
with a specification of the restrictions that a user can operate
within, such as which programs are allowed to be executed, which
resources they are allowed, and so on. The agent would then not
allow any access outside these limits. All of this allows for less
processing for the server, and allows for a fine tuning of user
rights to resources.