A glossary of agent community terms

Autonomy (in a software agent)

The ability to execute tasks without requiring user initiation, often without user confirmation, and sometimes even without user notification.

Blind signature

A means of cryptographically signing a message when its actual contents are not revealed to the person signing. See Schneier, page 93.

Denial of service

Often used in a security context, in which no information is compromised, but no one can get any work done, either. For example, a denial of service attack on a telephone system might involve large-scale destruction of cables or switching gear.

Dynamic clustering

Automatically forming groups or clusters of agents according to their momentary goals.

Firewall

A security perimeter of an organization's distributed computing environment. Usually, an organization which wishes to make certain aspects of, e.g., its Internet security easier will force all machines at the site to communicate through one firewall machine. The firewall itself is specially configured to reject most communications from the outside, and to filter the remainder according to various criteria. (Unfortunately, reliance on firewalling tends to mean that any misconfiguration of the firewall machine is fatal, because the rest of the organizations computers are probably very careless configured or monitored. It is nonetheless a common way for many companies on the Internet to try to protect themselves from outside crackers.)

Information filtering

The process of weeding out the information useful to some particular user from a larger stream. For example, email filtering or news filtering.

Information discovery

Actively seeking out new sources of information from locations of which the user may be unaware.

Island

An isolated group of agents that does not know about the existence of the larger whole. Possible behind firewalls, or accidentally.

Traffic analysis

From the perspective of a malicious third party, traffic analysis is the practice of determining who talks to whom, how often, and for how long. This information is often almost as useful as knowing the actual contents of the communication. For this reason, even if the conversation is encrypted, it is often the case that foiling traffic analysis (by hiding who it talking to whom) is still necesssary.

Trust (in a software agent)

Being sure that the agent will carry out the user's wishes correctly. This requires both that the agent know what it is that the user wishes done, and that the user knows that the agent knows it. If the user does not know what the agent is likely to do, he or she cannot give the agent very much autonomy in its actions.

Personalizability (in a software agent)

Being able to customize an agent such that it carries out tasks that are useful to its user, which may or may not resemble tasks carried out by other agents for other users. This customization could come from explicit programming, or from machine learning in which the agent infers user behavior from observation. For example, an email filtering agent should filter email according to what its user finds interesting.

Serendipitous matches

Serendipitous matches between users consist of matches made without the user necessarily intending to look for a match. A serendipitous match may involve a user who has not even thought to look for someone with similar interests on some topic, and may be surprised to find that there is any other user out there with such an interest.

Zero-knowledge proof

A cryptographic protocol in the interrogatee can convince an interrogator, solely by answering yes/no questions, the he or she knows something, but does not give the interrogator any information about what it is the interrogatee knows. For more information, consult Schneier, pages 85-87.