News Feed January-February 2024

Published papers

Publications accepted

• Zewei Shi, Ruoxi Sun, Jieshan Chen, Jiamou Sun, Minhui Xue. “The Invisible Game on the Internet: A Case Study of Decoding Deceptive Patterns”, accepted at The Web Conference 2024 (WWW, Core A*), short paper track. This paper introduces a comprehensive approach involving the interactions between the Adversary, Watchdog (e.g., detection tools), and Challengers (e.g., users) to formalize and decode deceptive pattern threats. Based on this, we propose a quantitative risk assessment system.
• Hongsheng Hu, Shuo Wang, Tian Dong, Minhui Xue. “Learn What You want To Unlearn: Unlearning Inversion Attacks against Machine Unlearning”, accepted at IEEE Symposium on Security and Privacy 2024 (Oakland, Core A*). This paper proposes unlearning inversion attacks on machine unlearning, showing that the unlearned data can be recovered through the differences of the original model and unlearned model.
• Sajal Halder, Michael Bewong, Arash Mahboubi, Yinhao Jiang, Rafiqul Islam, Zahidul Islam, Ryan Ip Ho Leung, Muhammad  Ejaz Ahmed, Gowri Ramachandran and Muhammad Ali Babar. ” Malicious Package Detection using Metadata Information”, Accepted at The Web Conference 2024 (WWW, Core A*).  The paper presents a model that utilizes metadata for malicious package detection, enhancing effectiveness and showcasing resilience against adversarial attacks.
• Yanjun Zhang, Ruoxi Sun, Liyue Shen, Guangdong Bai, Minhui Xue, Mark Huasong Meng, Xue Li, Ryan Ko, Surya Nepal. “Privacy-Preserving and Fairness-Aware Federated Learning for Critical Infrastructure Protection and Resilience”, accepted at The Web Conference 2024 (WWW, Core A*). The paper proposes Confined Gradient Descent (CGD) that enhances the privacy of federated learning by eliminating the sharing of global model parameters, allowing the proprietary confined models to adapt to the heterogeneity in federated learning, providing inherent benefits of fairness.
• Zihan Wang, Zhongkui Ma, Xinguo Feng, Ruoxi Sun, Hu Wang, Minhui Xue, Guangdong Bai. “CORELOCKER: Neuron-level Usage Control for AI Models”, accepted at IEEE Symposium on Security and Privacy 2024 (Oakland, Core A*). This paper proposes CoreLocker, employing the strategic extraction of a small subset of significant weights from the neural network, which serves as the access key to unlock the model’s complete capability, safeguarding the model from unauthorized use. 
• Nan Wu, Xin Yuan, Shuo Wang, Hongsheng Hu, Minhui Xue, “Cardinality Counting in “Alcatraz”: A Privacy-aware Federated Learning Approach”, in ACM The Web Conference 2024 (WWW ‘2024, Core A*). This paper proposes a privacy-preserving federated K-means clustering approach for cardinality counting in Smart Health.
• David D. Nguyen, David Liebowitz, Surya Nepal, Salil Kanhere, “Multiple Hypothesis Dropout: Estimating the Parameters of Multi-Modal Output Distributions” in 38th AAAI Conference on Artificial Intelligence (AAAI-24) (A*). The paper introduces a novel multiple-output function and investigates its applicability across supervised and unsupervised problems.
• Wai-Kong Lee, Raymond K. Zhao, Ron Steinfeld, Amin Sakzad, and Seong Oun Hwang “High Throughput Lattice-based Signatures on GPUs: Comparing Falcon and Mitaka“. IEEE Transactions on Parallel and Distributed Systems (TPDS), CORE A*. We develop the first high throughput implementation of the Falcon post-quantum digital signature scheme on Graphics Processing Units (GPUs). Falcon is one of the post-quantum digital signature scheme approved by the National Institute of Standards and Technology (NIST). With the fast deployment of GPU capabilities on cloud service platforms, our GPU-accelerated implementation may see adoptions in applications such as the Internet of Things (IoT).
• Zhi Zhang, Decheng Chen, Jiahao Qi, Yueqiang Cheng, Shijie Jiang, Yiyang Lin, Yansong Gao, Surya Nepal, Yi Zou, Jiliang Zhang, Yang Xiang, “SoK: Rowhammer on Commodity Operating Systems”, Accepted by AsiaCCS (Data61 target venue; CORE A; Acceptance Rate = 19%) In this paper, we systematize rowhammer attacks and defenses with a focus on DRAM-based commodity systems. Particularly, we have established a unified framework demonstrating how a rowhammer attack affects a commodity system.
• Chao Liu, Boxi Chen, Wei Shao, Chris Zhang, Kelvin Wong, Yi Zhang, “Unraveling Attacks in Machine Learning-based IoT Ecosystems: A Survey and the Open Libraries Behind Them” in IEEE Internet of Things (Target Venue, Q1, IF=10.6). Our study delves into the intricate security challenges posed by integrating Machine Learning (ML) with the Internet of Things (IoT). We thoroughly investigate a variety of security threats, including membership inference attacks, adversarial attacks, and poisoning attacks. Our research categorizes these threats according to adversary models, the objectives of the attacks, and the impacted security attributes, offering a depth of analysis beyond what previous studies have achieved. Furthermore, we scrutinize 65 different libraries designed to safeguard both model and data privacy within ML-powered IoT frameworks. This paper presents critical insights and proposes robust solutions aimed at enhancing the security of ML models and data amidst the rapidly evolving landscape of AI and IoT.
• Erik Buchholz, Sharif Abuadbba, Shuo Wang, Surya Nepal, and Salil S. Kanhere. “SoK: Can Trajectory Generation Combine Privacy and Utility?” accepted at Privacy Enhancing Technologies 2024 (PETS’24) Issue 3. [Data61 target venue; CORE A; Acceptance Rate (AR): 19.5%; Link to PDF]: This paper systematises approaches that target the privacy-preserving publication of location trajectories. Due to the prohibitive privacy-utility trade-off of conventional protection approaches, the release of synthetic data generated through generative deep learning models is considered a promising alternative. In a large-scale experimental study, both state-of-the-art models and generative models for sequential tasks from other domains are evaluated. The results highlight that currently, no solution can combine rigid privacy guarantees and good utility and that generative models from other domains are not readily applicable to trajectory generation. This work exposes a significant gap in the space of trajectory privacy and underlines the need for a generative trajectory model with strong privacy guarantees.
• Nan Wang, Dongxi Liu, “FlashSwift: A Configurable and More Efficient Range Proof With Transparent Setup”, in PETS 2024 (CORE A). The paper proposes a new zero-knowledge range proof by combining the techniques of Flashproof and SwiftRange with a transparent setup.
• Rayne Holland, Olya Ohrimenko, Anthony Wirth, “Single Round-trip Hierarchical ORAM via Succinct Indices,” ACM AsiaCCS 2024 (accepted, Core A). This paper presented a practical ORAM construction that matches the bandwidth overhead of the state-of-the-art approaches while reducing the client memory cost.  
• Yashothara Shanmugarasa, M. A. P. Chamikara, Hye-young Paik, Salil S. Kanhere, and Liming Zhu. “Local Differential Privacy for Smart Meter Data Sharing With Energy Disaggregation” accepted at The 20th International Conference on Distributed Computing in Smart Systems and the Internet of Things (DCOSS) (CORE B). This paper presents a novel LDP approach to facilitate the sharing of appliance-level energy consumption data over time while not revealing individual users’ appliance usage patterns.
• Diksha Goel, Hong Shen, Hui Tian, Mingyu Guo, “Effective Graph-Neural-Network based Models for Discovering Structural Hole Spanners in Large-Scale and Diverse Networks” in Expert Systems with Applications, Elsevier [Q1 Ranked Journal, Impact Factor: 8.5]. This paper designs GraphSHS and Meta-GraphSHS graph neural network models to efficiently discover structural hole spanner nodes in large-scale and diverse networks, achieving high accuracy and significantly reducing computational cost.
• Wei Shao, Yu Zhang, Pengfei Xiao, Kyle Kai Qin, Mohammad Saiedur Rahaman, Jeffrey Chan, Bin Guo, Andy Song, Flora D. Salim, “Transferrable contextual feature clusters for parking occupancy prediction“, in Pervasive Mobile Computing (Q1, IF=4.3). It provides a new solution for parking availability prediction when the local government does not have enough historical data.
• Hang Bui Thanh, Hamed Aboutorab, Arash Mahboubi, Yansong Gao, Nazatul Sultan, Aufeef Chauhan, Mohammad Zavid Parvez, Michael Bewong, Rafiqul Islam, Zahid Islam, Seyit A. Camtepe, Praveen Gauravaram, Dineshkumar Singh, M. Ali Babar, Shihao Yan, “Agriculture 4.0 and Beyond: Evaluating Cyber Threat Intelligence Sources and Techniques in Smart Farming Ecosystems” in Computer & Security, a Q1 ranked journal. This paper develops a comprehensive taxonomy of Cyber Threat Intelligence techniques and sources, specifically tailored for the Smart Farming Infrastructure context, addressing the unique cyber threat challenges in this domain.
• Iqbal H Sarker, Helge Janicke, Mohamed Amine Ferrag, Sharif Abuadbba, “Multi-aspect rule-based AI: Methods, taxonomy, challenges and directions toward automation, intelligence and transparent cybersecurity modeling for critical infrastructures” Accepted to Elsevier Internet of Things 2024, a Q1 ranked journal. This article provides an in-depth study on multi-aspect rule-based AI modeling considering human interpretable decisions as well as security automation and intelligence for critical infrastructures. It also provides a taxonomy of rule generation methods by considering not only knowledge-driven approaches based on human expertise but also data-driven ones, i.e., extracting insights or useful knowledge from data, and their hybridization.
• Nazatul H. Sultan, Shabnam Kasra Kermanshahi, Hong Yen Tran, Shangqi Lai, Vijay Varadharajan, Surya Nepal, Xun Yi, “Securely Sharing Outsourced IoT Data: A Secure Access and Privacy Preserving Keyword Search Scheme”, in the Ad Hoc Networks, Elsevier (Q1 Rank). This paper proposes a novel keyword search scheme using computationally lightweight cryptographic primitives to empower the IoT data owners to securely share, store and manage encrypted data in the CSPs, providing better security and privacy.
• Marthie Grobler and Tooba Aamir had a chapter accepted in the book ‘Psybersecurity: Human factors of cyber defence’. The chapter is titled ‘Building cognitive resilience for enhanced cyber governance’. In this chapter, we examine how cybersecurity attacks and events aim to exploit human psychology in the context of governance, as well as explore the triggers that adversaries manipulate to breach security. We discuss how these tactics compromise decision-making processes, data sharing, and citizen engagement, and how this would contribute to the potential erosion of public trust, compromise of sensitive data, and disruptions in governance operations. We further consider to what extent cyber resilience would contribute to one person or entity being more susceptible to cyber incidents than others and explore how cognitive resilience can build a cyber immunity that can be adopted in modern cyber governance frameworks.
• M. Sayat; B. Shajilal; S. P. Kish; S. M. Assad; T. Symul; P. K. Lam; N. Rattenbury; J. Cater, “Satellite-to-Ground Continuous Variable Quantum Key Distribution: The Gaussian and Discrete Modulated Protocols in Low Earth Orbit,” in IEEE Transactions on Communications, doi: 10.1109/TCOMM.2024.3359295. The Gaussian modulated continuous variable quantum key distribution (GM-CVQKD) protocol is known to maximise the mutual information between two parties during quantum key distribution (QKD). An alternative modulation scheme is the discrete modulated CVQKD (DM-CVQKD) protocol. In this paper, we study the Phase Shift Keying (M-PSK) and Quadrature Amplitude Modulation (M QAM) DM-CVQKD protocols along with the GM-CVQKD protocol over a satellite-to-ground link in the low SNR regime. We use a satellite-to-ground link model which takes into account geometric losses, scintillation, and scattering losses from the link distance, atmospheric turbulence, and atmospheric aerosols, respectively. In addition, recent multidimensional (MD) and multilevel coding and multistage decoding (MLC-MSD) reconciliation method models in combination with multiedge-type low-density parity-check (MET-LDPC) code models have been used to determine the reconciliation efficiency. The results show that GM-CVQKD outperforms DM-CVQKD. In addition, GM-CVQKD with MD reconciliation outperforms GM-CVQKD with MLC-MSD reconciliation in the finite size limit by producing positive secret key rates at larger link distances and lower elevation angles.
• Cody Christopher, Alban Grastien (2023). “ Critical Observations in Model-Based Diagnosis“ to Journal of Artificial Intelligence (AIJ), Elsevier (Q1 – CS 11.4, IF 14.4), the peak body journal for AI, pending minor revisions. Co-authored with former PhD supervisor formerly of the ANU, this paper is a condensed and updated version of material presented in PhD Thesis (2019), in AI Explanability extensions for classical AI diagnosis algorithms in formal systems.