Toward Adversarial Online Learning and the Science of Deceptive Machines

Intelligent systems rely on pattern recognition and signaturebasedapproaches for a wide range of sensors enhancing situationalawareness. For example, autonomous systems dependon environmental sensors to perform their tasks and securesystems depend on anomaly detection methods. The availabilityof lar...

Full description

Saved in:
Bibliographic Details
Main Author: Abramson,Myriam
Format: Report
Language:English
Online Access:Request full text
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:Intelligent systems rely on pattern recognition and signaturebasedapproaches for a wide range of sensors enhancing situationalawareness. For example, autonomous systems dependon environmental sensors to perform their tasks and securesystems depend on anomaly detection methods. The availabilityof large amount of data requires the processing of datain a streaming fashion with online algorithms. Yet, just asonline learning can enhance adaptability to a non-stationaryenvironment, it introduces vulnerabilities that can be manipulatedby adversaries to achieve their goals while evadingdetection. Although human intelligence might have evolvedfrom social interactions, machine intelligence has evolved asa human intelligence artifact and been kept isolated to avoidethical dilemmas. As our adversaries become sophisticated,it might be time to revisit this question and examine how wecan combine online learning and reasoning leading to the scienceof deceptive and counter-deceptive machines. AAAI Fall Symposium , 12 Nov 2015, 14 Nov 2015,