Trustworthy Intelligent Machine Engineering

Securing the Machine Economy Future

The last decade has witnessed the proliferation of data collection and artificial intelligence in our daily lives. Technological advances under the Internet of Things moniker have extended the scope of data collection beyond our interactions with the digital world to cover real-world phenomena. This data fuels artificial intelligence systems that decide the content of our mailboxes and our newsfeeds. Artificial intelligence has also been permeating our physical world. Augmented by sensors, antennas, and microprocessors, many devices in our daily lives are now able to sense their surroundings, interact with cloud services, and respond to commands issued over the Internet.

The next evolution would emerge when intelligent machines can coordinate to carry out complex processes with minimal human involvement. The seeds of this evolution have already been planted in machine-to-machine data marketplaces and smart grids. Via cooperation, machines can automate many low value-added activities, freeing humans to focus on hard-to-automate decisions. For instance, a personal weather station can automatically negotiate and exchange its data for network bandwidth, electricity, and maintenance, based on some parameters set by its owner. Correctly done, such a machine economy can bring about tremendous productivity and profound impacts on our lives.

The machine economy hinges on integration and security. The integration needs to happen across organizational and jurisdictional boundaries, as any non-trivial process of intelligent machines would not be bounded within an organization or even country. Moreover, artificial intelligence models must also evolve at a higher pace in response to the volume and velocity of data in the machine economy. Such pace demands a seamless integration between data sources, model development pipelines, and production environments. Security would be paramount in such tight integration because any tampering of data, models, or decisions would have cascading effects that can impact the lives of many people.

Traditionally, the integration happens via intermediaries such as If This Then That (IFTTT), which resolve discrepancies between services' interfaces and coordinate them according to predefined scenarios. This approach is unsustainable in the machine economy, as a few orchestration services cannot scale to the global demand and represent single points of failure. Moreover, private ownership and jurisdictional boundaries of orchestration services are contentious with the cross-organizational and international nature of the machine economy. Finally, the concealed nature of these services means that the security of the entire machine economy hinges on the belief that the service providers would secure the data and the integration.

A better approach would be to let intelligent machines coordinate themselves and make the process transparent, traceable, and verifiable.

Distributed Ledger Engineering Research

Over the last decade, the Distributed Ledger technology (DLT), also known as Blockchain, has emerged as a means for a group of rational agents to connect and coordinate with each other to carry out a process without relying on an opaque and centralized intermediary. This technology was used to create scarcity in a digital world, where everything can be copied and clone. This scarcity has enabled the so-called cryptocurrencies, such as Bitcoin and Ethereum, as well as digital collectibles such as Cryptokitties. In recent years, this technology has been increasingly used in the banking and financial sector to enable business processes, such as international Letter of Credit issuance. Beyond currency and finance, DLT has been applied in government services such as notary and providing proof-of-existence of documents. These usages show that DLT represents a blueprint for a new way of orchestrating intelligent machines and services in a decentralized, transparent, and verifiable way.

Distributed Ledgers do not exist in vacuum. Instead, in any usage scenarios that involve DL, they are in constant contact and interaction with other software systems, services, and intelligent machines. Thus, whether DL can make an effective solution depends a great deal on how it is integrated with other systems. Moreover, as DL is a software system itself, the successful utilization of DL also depends on how it is designed and engineered. As any emerging technologies, DLT has its terminologies, characteristics, and constraints that software engineers must carefully consider in order to utilize it effectively.

Research Areas

The goal of my research is to support industries, governments, and societies in leveraging DLT to address current challenges and prepare for the emerging machine economy. The foundation of my research and practice is evidence-based software engineering research methods, based on inputs from scientific literature, empirical data, and experimentation. Upon this foundation, our research addresses three topic areas.

Techniques and Frameworks for DLT System Engineering

The first area of research concerns with the development, cataloguing, and evaluating techniques and frameworks for architecting and engineering DLT systems. Topics of interest in this area include:

• Development of a Reference Architecture of Holistic DLT systems
• Discovering and cataloguing architectural tactics and patterns for DLT systems
• Visualisation and documentation of a DLT systems.

On-going Projects:

• An approach for architecting DLT systems using state machines: This project aims to propose, specify, and evaluate a novel approach to architect DLT systems using the concepts of state machines. Outcomes of this project include detailed specifications of the approach and empirical data concerning the employment of the approach in real projects.
• A review or architectural design techniques for DLT systems: This project aims to systematically identify and review the state of the art on architectural models, tactics, and patterns for designing DLT systems.
• An approach to develop DLT networks based on prototyping: This project aims to propose, specify, and evaluate an approach to architect and deploy DLT networks, which relies on rapid prototyping to guide the design process. Outcomes of this project include detailed specifications of the approach and results from case studies where the approach has been applied.

Tools and Automation Support for DLT System Engineering

The second area of research concerns with the development, cataloguing, and evaluation of tools and automations to support engineering of DLT systems. Topics of interest include:

• Design and development of platforms for rapid prototyping of DLT networks
• Development of model-driven approaches to DLT system engineering
• Development of data-driven recommendation mechanisms for DLT system design

On-going Projects:

• Platform for automating the generation and evaluation of DLT network prototype: This project aims to develop, deploy, and evaluate a platform for rapid prototyping and evaluation of DLT network prototypes to support the architecting of DLT network element of the system. The outcome of this project would be a fully functional proof-of-concept of the platform deployed at the University of Adelaide, along with detailed design specification of the system.

Tools and Techniques for Secure-by-Design DLT System

The third area of research concerns with techniques and tools for facilitating secure-by-design of DLT. We focus on a holistic perspective that considers not just the ledger networks but the interfaces between the ledger and other system components. Some topics of interest include:

• Establishing a security framework for the interface between ledger and the rest of the system
• Assessing the security of multi-ledger architecture
• Assessing the trade-off between security and privacy in a DLT system.

Application Areas

Distributed Ledger Engineering Research aims to support industries, governments, and societies to address their current challenges as well as prepare themselves for the emerging machine economy. My research program target three application areas as follow.

Cross-organization Processes and Provenance

Frameworks, techniques, and tools developed by my research lay a foundation for the development of DLT systems that track entities moving between organizations and manage multi-party processes. Some particular application areas include:

• Security patch management
• Logistic management of commodity
• Hardware inventory management
• Artificial intelligence provenance

On-going Projects:

• Accountable AI with DLT: This project aims to develop, deploy, and evaluate a mechanism for tracking machine learning pipelines to create immutable records of AI models throughout its life cycle. The outcome of this project includes a functional proof-of-concept and detailed design specification of the system.

Trustworthy Data and Processes at the Tactical Edge

Technological advances under the Internet of Things (IoT) moniker have made the deployment of sensors and autonomous systems desirable and feasible. Located at the tactical edge, these systems often operate in congested and potentially contested environments. Therefore, they require novel mechanisms to coordinate themselves and ensure the trustworthiness of the data and instructions that they receive, without constant interaction with remote servers. Frameworks, techniques, and tools developed by my research lay a foundation for developing DLT systems that can operate at the tactical edge to coordinate these edge systems. Some particular application areas include:

• Edge-based DLT systems for tactical edge operation
• Integrity assurance of data collected from the edge

On-going Projects:

• Distributed Ledger at Tactical Edge: This project aims to evaluate the feasibility and design the mechanisms for deploying DLT at the tactical edge, to provide a platform for trustworthy data collection and processing among autonomous edge devices. The outcomes of this project includes prototypes of DLT operating at the edge on low-power devices, benchmarking results of these DLT, and supporting tools and mechanisms for deploying these DLT systems.

Management of Shared Resources over Open Networks

Data and computing infrastructure are some of the most prominent resources that fuel modern research activities. Due to the scarcity of these resources, sharing among research institutes, industry partners, and governments is desirable and beneficial. This sharing problem is a case of cross-organizational process management over open networks. Frameworks, techniques, and tools developed by my research lay a foundation for developing DLT systems that enable these scenarios. Some particular application areas include:

• Decentralized testbed for IoT research
• Platform for rapid prototyping and evaluation of DLT systems

On-going Projects:

• Decentralized IoT Testbed (DTB): This project aims at to develop, deploy, and evaluate a decentralised platform for conducting experiments on IoT device kits. This platform maintains traceability of IoT devices making up the device kits and results of the experiments conducting on them. DLT enables the decentralized ownership and management of this sensitive information, allowing multiple organizations to participate and operate the testbed to provide services to researchers and developers. The outcome of this project would be a fully functional proof-of-concept of the testbed deployed at the University of Adelaide, along with detailed design specification of the system.