Privacy and security of cognitive augmentation in policing
The rapid technological advancements of recent years have had a tremendous and multifaceted impact on society, leading to innovations such as the Internet of Things (IoT) and the advent of intelligent environments. Through these new intelligent environments, continuous streams of telemetry data are recorded, allowing users to remotely monitor and adjust their devices and enabling the automatic invocation of sensor/actuator processes, for on-the-fly adjustment of environmental conditions. This, coupled with research developments in the field of Artificial Intelligence (AI), open the way for new and innovative ways through which users and intelligent environments could interact. One such human-environment interaction method that has gained some traction in recent years is automatic cognitive-based interface, where brain-waves are captured through specialised hardware, processed and translated in formats that can be processed by controller, sensor, actuator and computer systems.
This PhD thesis focuses on the development of cognitive privacy and crime investigation techniques using deep learning models. The research aims to address three major challenges: (1) designing process capabilities to cater for dynamic environments, (2) ensuring privacy preservation during sensor-derived EEG data, and (3) augmenting crime investigation process and transitioning from the existing manual paradigm to automated AI-powered data analytics. The proposed solutions involve combining sequence-based and convolution-based deep learning models with statistical measures to obfuscate the identity of data owners, prevent profiling, and improve the detection of brain activity and crime type detection.
The first contribution introduces a cognitive augmentation framework, which identifies the relevant research and technologies, and explains its application amidst real-world use cases in policing. The second contribution proposes AI-enabled EEG model, namely Cognitive Privacy to safeguard data and identifies users and their tasks from EEG data. The data is protected from disclosure using normalized correlation analysis and classifies subjects and their tasks using a long-short term memory (LSTM) deep learning technique. The third contribution presents a new Cognitive Computing enabled Convolution Neural Network (CC-CNN) model, that allows the classification of incidents into crime categories and their associated criminal acts. The proposed CC-CNN model can be used by investigators to gain a better understanding of crimes by processing written statements and witness accounts.
The cognitive augmentation framework proposed in this research is a significant contribution to academic literature, as it provides a comprehensive approach to integrating cognitive technologies into real-world use cases, bridging the gap between research and practice. The AI-enabled EEG-based model proposed in this research, called Cognitive Privacy, is significant from an academic perspective as it presents a novel approach to safeguarding the privacy and security of brain activity data, while accurately classifying subjects and their tasks from EEG data. This has significant implications for various industries beyond policing, including healthcare and finance. The significance of the CC-CNN model, from an academic perspective, is determined by its novel approach to classifying incidents into crime categories and their associated criminal acts using written statements and witness accounts. The model’s proposed application in the field of policing offers promising results and adds to the growing body of research on the use of cognitive technologies in law enforcement. The contributions of this thesis not only offer a road-map for augmenting human capabilities through AI, but also provide insight into the potential ethical and practical ramifications of a potential future reliance on AI-powered systems for human society.
