Bruno N. Di Stefano

Computational intelligence based architecture for cognitive agents

Abstract We discuss some limitations of reflexive agents to motivate the need to develop cognitive agents and propose a hierarchical, layered, architecture for cognitive agents. Our examples often involve the discussion of cognitive agents in highway traffic models. A cognitive agent is an agent capable of performing cognitive acts, i.e. a sequence of the following activities: “ Perceiving ” information in the environment and provided by other agents, “ Reasoning ” about this information using existing knowledge, “ Judging ” the obtained information using existing knowledge, “ Responding ” to other cognitive agents or to the external environment, as it may be required, and “ Learning ”, i.e. changing (and, hopefully augmenting) the existing knowledge if the newly acquired information allows it. We describe how computational intelligence techniques (e.g., fuzzy logic, neural networks, genetic algorithms, etc) allow mimicking to a certain extent the cognitive acts performed by human beings. The order with which the cognitive actions take place is important and so is the order with which the various computational intelligence techniques are applied. We believe that a hierarchical layered model should be defined for the generic cognitive agents in a style akin to the hierarchical OSI 7 layer model used in data communication. We outline in broad sense such a reference model.

OSI Network‐layer Abstraction: Analysis of Simulation Dynamics and Performance Indicators

The Open Systems Interconnection (OSI) reference model provides a conceptual framework for communication among computers in a data communication network. The Network Layer of this model is responsible for the routing and forwarding of packets of data. We investigate the OSI Network Layer and develop an abstraction suitable for the study of various network performance indicators, e.g. throughput, average packet delay, average packet speed, average packet path‐length, etc. We investigate how the network dynamics and the network performance indicators are affected by various routing algorithms and by the addition of randomly generated links into a regular network connection topology of fixed size. We observe that the network dynamics is not simply the sum of effects resulting from adding individual links to the connection topology but rather is governed nonlinearly by the complex interactions caused by the existence of all randomly added and already existing links in the network. Data for our study was gathe...

Creature Learning to Cross a CA Simulated Road

Agent-based models approximate the behaviour of simple natural and man-made systems. We present a simple cognitive agent capable of evaluating if a strategy has been applied successfully and capable of applying this strategy again with small changes to a similar but new situation. We describe some experimental results, present our conclusions, and outlines future work.

Cognitive agents: Functionality & performance requirements and a proposed software architecture

We present a short review of the concepts of agent and cognitive agent. Our examples often involve agents in highway traffic models. We examine functionality & performance requirements of cognitive agents. We propose the architecture of an application independent software implementation of a generic cognitive agent able of providing the required functionality and performance. We suggest various computational intelligence methodologies for giving the proposed cognitive agent learning abilities.

Simulated Naïve Creature Crossing a Highway

We present a simple cognitive agent, a “Simulated Naive Creature”, capable of evaluating if a strategy has been applied successfully. The agents are born as “tabula rasa”, i.e. a “blank slate”. They are provided with a mechanism to reason and plan toward their goal, but they have no built-in knowledge-base of their environment. Our simulation shows that the performance of the agents is affected by the conditions of the environment (e.g. the traffic density on the highway and by the crossing point location on the highway) and by their fears and desires.

Naïve creature learns to cross a highway in a simulated CA-like environment

We present a model of simple cognitive agents, called “creatures”, and their learning process, a type of “social observational learning”, that is each creature learns from the behaviour of other creatures. The creatures may experience fear and/or desire, and are capable of evaluating if a strategy has been applied successfully and of applying this strategy again with small changes to a similar but new situation. The creatures are born as “tabula rasa”; i.e. without built-in knowledge base of their environment and as they learn they build this knowledge base. We study learning outcomes of a population of such creatures when they are learning how to safely cross various types of highways. The highways are implemented as a modified Nagel-Schreckenberg model, a CA based highway model, and each creature is provided with mechanism to reason to cross safely the highway. We present selected simulation results and their analysis.

Software Implementation of Population of Cognitive Agents Learning to Cross a Highway

We describe the model and the software implementation of population of simple cognitive agents, naive creatures experiencing fear and/or desire while learning to cross a highway. The creatures use an observational learning mechanism for adoption or rejection of a strategy to cross the highway. Presented simulation results are consistent with the fact that crossing a highway becomes more difficult with increase of cars density and it is affected by the creatures’ fears and desires. The transfer the knowledge base acquired in one environment to another one combined with creatures ability to change a crossing point improves creatures success of crossing a highway.

Entropy Based Detection of DDoS Attacks in Packet Switching Network Models

Distributed denial-of-service (DDoS) attacks are network-wide attacks that cannot be detected or stopped easily. They affect “natural” spatio-temporal packet traffic patterns, i.e. “natural distributions” of packets passing through the routers. Thus, they affect “natural” information entropy profiles, a sort of “fingerprints”, of normal packet traffic. We study if by monitoring information entropy of packet traffic through selected routers one may detect DDoS attacks or anomalous packet traffic in packet switching network (PSN) models. Our simulations show that the considered DDoS attacks of “ping” type cause shifts in information entropy profiles of packet traffic monitored even at small sets of routers and that it is easier to detect these shifts if static routing is used instead of dynamic routing. Thus, network-wide monitoring of information entropy of packet traffic at properly selected routers may provide means for detecting DDoS attacks and other anomalous packet traffics.

Detection & study of DDoS attacks via entropy in data network models

We detect & study packet traffic anomalies similar to DDoS attacks using information entropy. We perform network-wide monitoring of information entropy of packet traffic at a small number of selected routers. Our method is based on the fact that DDoS attacks change the “natural” order and randomness of packet traffic passing through monitored routers when an attack is taking place in the network. Through this change we detect the start of the attack and study its evolution. We conduct this study for packet-switching networks using static and dynamic routing.

Stochastic Model of Cognitive Agents Learning to Cross a Highway

We describe a stochastic model of simple cognitive agents (“creatures”) learning to cross a highway. The creatures are capable of experiencing fear and/or desire to cross and they use an observational learning mechanism. Our simulation results are consistent with real life observations and are affected by the creatures’ fears and desires, and the conditions of the environment. The transfer of the knowledge base acquired by creatures in one environment to the creatures operating in another one improves creatures’ success of crossing a highway.