A SECURITY ARCHITECTURE FOR AGENT COMMUNICATION LANGUAGES
 
			    James Mayfield
		     The Johns Hopkins University
		      Applied Physics Laboratory
		     Laurel, MD  20723--6099  USA
		      James.Mayfield@jhuapl.edu

			      Tim Finin
	     Computer Science and Electrical Engineering
	       University of Maryland Baltimore County
		      Baltimore, MD  21250  USA
			  finin@cs.umbc.edu

			     January 1996


One of the essential features of a software agent is its ability to
cooperate with other software agents.  This cooperation requires, in
general, that software agents be able to communicate in an
appropriately rich agent communication language (ACL) and associated
protocols.  For an ACL to be effective in an open environment like the
Internet, it must support security, privacy, the integrity of data,
and authentication of agent identity.  We discuss some basic and
extended security requirements for software agents and an architecture
to satisfy those requirements for KQML-speaking agents.  Many of these
security features will be provided by transport mechanisms which carry
the ACL (e.g., sockets, HTTP, SMTP).  However, security properties
must be part of and reflected in the ACL model and cannot simply be
relegated to the lower levels of the communication protocol stack.