Deceptive Autonomous Agents

Date

2020-01-09 10:31

Supervisor/s

Journal Title

Journal ISSN

Volume Title

Publisher

Cranfield University

Department

Type

Presentation

ISSN

Format

Free to read from

Citation

Sarkadi, Stefan (2020). Deceptive Autonomous Agents. Cranfield Online Research Data (CORD). Presentation. https://doi.org/10.17862/cranfield.rd.11558397.v1

Abstract

Recent advances in Artificial Intelligence (AI) along with recent events revolving around the problem of fake news indicate new and critical potential threats to intelligence analysis, defence, security, and, by extension, to modern society in general. One such threat that we can derive from the development of AI is the emergence of malicious autonomous artificial agents that could develop their own reasons and strategies to act dishonestly. In order to be able to prevent or mitigate the malicious behaviour of deceptive artificial and autonomous agents, we must first understand how they might be designed, modelled, or engineered. In this work, we aim to model and study how artificial agents that deceive and detect deception can be engineered, as well as how such agents might impact the common good.

Description

Software Description

Software Language

Github

Keywords

'Deception', 'Malicious AI', 'Complex Reasoning', 'DSDS19', 'DSDS19 Technical Paper', 'Artificial Intelligence and Image Processing not elsewhere classified'

DOI

10.17862/cranfield.rd.11558397.v1

Rights

CC BY-NC 4.0

Relationships

Relationships

Supplements

Funder/s

KCL PhD Funding

Collections