Computer Science

Permanent URI for this collection

Browse

Recent Submissions

Now showing 1 - 20 of 85
  • ItemOpen Access
    Gaze-Contingent Multi-Modal and Multi-Sensory Applications
    (2024-11-07) Vinnikov, Margarita; Allison, Robert
    Gaze-contingent displays are applications that are driven by the user's gaze. They are an important tool for many multi-modal and multi-sensory domains. They can be used to precisely control the retinal image in real time to study visual control of natural behaviour through experimentation, or to improve user experience in virtual reality applications. In this dissertation, I explored the application of gaze-contingent display technology to dierent models and senses and evaluate whether such applications can be useful for simulation, psychophysical research and human-computer interaction. Specically, I have looked at the visual gaze-contingent display and an audio gaze-contingent display. I examined the effects of simulated visual defects on user's perception and control of self-motion during locomotion. I found that gaze-contingent display simulations of visual defects signicantly altered visual patterns and impaired the accuracy and precision of judgement of heading. I also examined the impact of simulating gaze-contingent depth-of-field for monocular and stereoscopic displays. The experimental data showed that the alleviation of negative eects associated with stereo displays depends on the user's age and the types of scenes that are viewed. Finally, I simulated gaze-contingent audio displays that imitated the cocktail party effect. My audio enhancement techniques turned to be very benecial for applications that have to deal with user's attention to multiple sources of sounds (speakers) such as teleconferences and social games. Finally, in this dissertation, I demonstrated that gaze-contingent systems can be used in many aspects of virtual system design and if combined together (used for multiple cues and senses) can be a very powerful tool for augmenting and improving the overall user experience.
  • ItemOpen Access
    Leveraging the Power of Images: Image Recommendation to Enhance Issue Reports
    (2024-11-07) Tan, Xuchen; Nayebi, Maleknaz
    The trend of sharing images and image-based social networks has eventually changed the landscape of social networks. As a result, this shift has impacted social coding platforms, and previous studies showed that image sharing has become increasingly popular among software developers. However, most developers’ productivity assistance tools predominantly rely on textual content only. To enhance issue reports, this study focuses on three primary objectives: (i) identifying issue reports that benefit from image sharing and processing in Bugzilla, (ii) identifying the type of image that would improve the bug report, and (iii) conducting a comprehensive qualitative and quantitative evaluation of the tool’s performance and impact. The quantitative evaluation demonstrates that our tool achieves an average recall of 78% and an average F1-score of 74% in predicting the necessity of including image attachments in issue reports. Moreover, our qualitative evaluation of software developers showed that 75% of the developers found the overall design and recommendations of our method practically useful for issue reporting. This study, along with its associated dataset and methodology, represents the first research on recommending images to developers for enhanced issue report communication. Our results illuminate a promising trajectory for enhanced and visual productivity tools for developers.
  • ItemOpen Access
    Empirical Analysis and Enhancement of Machine Learning Software Documentation
    (2024-11-07) Sharuka Promodya Thirimanne; Nayebi, Maleknaz; Datta, Suprakash
    As machine learning gains popularity, individuals from diverse fields and skill levels integrate it into their workflows. However, many lack software engineering experience, impacting the usability of documentation. Additionally, the current machine learning documentation and its issues are insufficiently addressed in the literature. This thesis comprises two papers. In the first paper, we compared the content and design differences of TensorFlow tutorials and analyzed the profiles of users who asked questions about them. We also developed a comprehensive taxonomy of TensorFlow documentation issues. In the second paper, we examined the potential of leveraging generative AI to augment machine learning documentation. We proposed a method to augment TensorFlow API documentation by addressing documentation-related questions using large language models. This thesis highlights the need for machine learning documentation to accommodate diverse skill levels as its use expands across domains and showcases the potential of generative AI to automate documentation augmentation.
  • ItemOpen Access
    Anonymity in Developer Communities: Insights from Developer Perceptions and Stack Overflow Profiles
    (2024-11-07) Lemango, Elim Yoseph; Nayebi, Maleknaz
    This thesis consists of two studies: an interview study with 34 early-career developers and a mining study analyzing 130,000 developer profiles. The interview study examines developers' definitions of anonymity, their preferences for anonymity, and their engagement with privacy policies. It also explores whether presenting privacy policies using contextual integrity principles improves understanding of the privacy policies. The developers from the interview study defined anonymity as withholding identifiable information like name, location, and professional background. The mining study investigates how much information developers share across platforms and the ease of retrieving their professional profiles. Our findings show that using Stack Overflow location and screen name in LinkedIn searches narrows down profiles, but cross-linking Stack Overflow profile data information with GitHub or Twitter adds noise. This research provides valuable insights into how developers define anonymity, and how that affects their behaviour when using social coding platforms.
  • ItemOpen Access
    Image White Balance for Multi-Illuminant Scenes
    (2024-11-07) Arora, Aditya; Derpanis, Konstantinos G.
    Performing white-balance (WB) correction for scenes with multiple illuminants remains a challenging task in computer vision. Most previous methods estimate per-pixel scene illumination directly in the RAW sensor image space. Recent work explored an alternative fusion strategy, where a neural network fuses multiple white-balanced versions of the input image processed to sRGB using pre-defined white-balance settings. Inspired by this line of work, we present two contributions targeting fusion-based multi-illuminant WB correction. First, we introduce a large-scale multi-illumination dataset rendered from RAW images to support training fusion models and evaluation. The dataset comprises over 16,000 sRGB images with ground truth sRGB white-balance corrected images. Next, we introduce an attention-based architecture to fuse five white-balance settings. This architecture yields an improvement of up to 25% over prior work.
  • ItemOpen Access
    Symmetry-based monocular 3D vehicle ground-truthing for traffic analytics
    (2024-11-07) Tran, Trong Thao; Elder, James
    3D object detection is critical for autonomous driving and traffic analytics. Current research relies on LiDAR-derived ground truth for training and evaluation. However, LiDAR ground truth is expensive and usually inaccurate in the far field due to sparse LiDAR returns. Assuming a fully calibrated camera and a 3D terrain model, we explore whether inexpensive RGB imagery can be used to obtain 3D ground truth based on the bilateral symmetry of motor vehicles. From manually annotated symmetry points and tire-ground contact points, we infer a vertical symmetry plane and 3D point cloud to estimate vehicle location, pose, and dimensions. These estimates are input into a probabilistic model derived from a standard public motor vehicle dataset to form maximum a posteriori estimates of remaining dimensions. Evaluations on a public traffic dataset show that this novel symmetry-based approach is more accurate than LiDAR-based ground-truthing on single frames and comparable to LiDAR-based methods that propagate information across frames.
  • ItemOpen Access
    An Axiomatic Perspective on Anomaly Detection
    (2024-11-07) Wyke, Chester Samuel; Urner, Ruth
    A major challenge for both theoretical treatment and practical application of unsupervised learning tasks, such as clustering, anomaly detection or generative modeling, is the inherent lack of quantifiable objectives. Choosing methods and evaluating outcomes is then often a matter of ad-hoc heuristics or personal taste. Anomaly detection is often employed as a preprocessing step to other learning tasks, and unsound decisions for this task may thus have far-reaching consequences. In this work, we propose an axiomatic framework for analyzing behaviours of anomaly detection methods. We propose a basic set of desirable properties (or axioms) for distance-based anomaly detection methods and identify dependencies and (in-)consistencies between subsets of these. In addition, we include empirical results, which demonstrate the benefits of this axiomatic perspective on behaviours of anomaly detection methods. Our experiments illustrate how some commonly employed algorithms violate, perhaps unexpectedly, a basic desirable property. Namely, we highlight a material problem with a commonly used method called Isolation Forest, related to infinite bands of space likely to be labelled as inliers that extend infinitely far away from the training data. Additionally, we experimentally demonstrate that another common method, Local Outlier Factor, is vulnerable to adversarial data poisoning. To conduct these experimental evaluations, a tool for dataset generation, experimentation and visualization was built, which is an additional contribution of this work.
  • ItemOpen Access
    Neural Document Segmentation Using Weighted Sliding Windows with Transformer Encoders
    (2024-11-07) Abbasi, Saeed; An, Aijun; Davoudi, Heidar
    The subdivision of documents into semantically coherent segments is a fundamental challenge in Natural Language Processing (NLP), with notable applications in information retrieval and question answering. Effective text segmentation is crucial for enhancing Retrieval-Augmented Generation (RAG) systems by providing coherent segments that improve the contextual accuracy of responses. We introduce a weighted sliding window framework, WeSWin, that effectively segments arbitrarily long documents using Transformers. WeSWin consists of overlapping document partitioning followed by the weighted aggregation of multiple sentence predictions within the generated windows, ensuring that sentences with larger context visibility contribute more to the ultimate label. Additionally, we propose a multi-task training framework, WeSWin-Ret, which combines text segmentation with an auxiliary sentence retrieval task. This approach injects query-awareness into the embedding space of the shared Transformer, resulting in improved segmentation performance. Extensive experiments demonstrate that our methods outperform state-of-the-art approaches. On the Wiki-727k benchmark, both our WeSWin and WeSWin-Ret models surpass existing works based on BERT, RoBERTa, or LongFormer Transformers. Notably, our RoBERTa baseline often matches LongFormer’s performance while being significantly more efficient in training and inference. We validate our model’s robustness on domain-specific segmentation benchmarks, including en_city, en_disease, and an industrial automotive dataset, demonstrating generalizability across domains. Lastly, our model proves to be highly effective in enhancing down-stream RAG applications by providing cohesive chunks for knowledge retrieval.
  • ItemOpen Access
    IoT network Malicious Behaviour Profiling Based on Explainable AI Using LSTM and SHAP
    (2024-11-07) Niktabe, Sepideh; Habibi Lashkari, Arash
    The proliferation of IoT devices has enhanced connectivity but exposed networks to new cyber threats, particularly from botnets. Detecting and identifying malicious data is critical for early threat detection, understanding botnet attack patterns, and deploying countermeasures. This research proposes an IoT Bot detection and identification profiling model using XAI. The proposed model introduces a novel feature selection techqnique with the XGBoost algorithm and a correlation-based feature selection technique to enhance efficiency. An optimized LSTM neural network enables accurate bot detection and identification, with hyperparameters selected using the Bayesian Optimization algorithm. SHAP analysis provides insightful individual and collective bot characteristic profiles. The model’s performance was evaluated using the augmented BCCC-Aposemat-Bot-IoT-24 dataset, built upon the Aposemat-Bot-IoT-23 dataset, and compared against established models assessed primarily on the same dataset in previous research. The results showed that the proposed model performed comparably to these models, with distinct advantages, including handling sequential and time-series data, managing imbalanced datasets, and providing explainable insights into botnet behavior. The model’s design also emphasizes computational efficiency, making it potentially suitable for deployment in resource-constrained environments.
  • ItemOpen Access
    Robust Representation Learning Solutions for Wireless Sensing Applications
    (2024-11-07) Barahimi, Borna; Tabassum, Hina
    WiFi sensing, a technique for utilizing wireless signals for monitoring human activities and environmental conditions, holds substantial potential in diverse applications including human activity recognition (HAR). It offers a powerful, continuous, and non-intrusive monitoring solution. This technology eliminates the need for wearable sensors, and even functions outside the line-of-sight. However, the large-scale deployment of WiFi sensing faces several challenges: (1) limited computational power in WiFi devices, (2) the cost and complexity of annotating channel state information (CSI) data, and (3) ensuring model generalization across different environments. The first part of the thesis addresses the limited computation power of edge devices by developing a Real-time Sensing and Compression Network (RSCNet). RSCNet is a cloud-based architecture designed to alleviate computational constraints on edge devices. It achieves this through efficient CSI compression at the edge and subsequent sensing and reconstruction in the cloud. RSCNet employs window-based CSI compression and LSTM-based recurrent blocks, significantly reducing computational demands and communication overheads while maintaining high sensing accuracy. The second part of the thesis addresses the issue of limited labeled data by developing self-supervised learning (SSL) method, namely Context-Aware Predictive Coding (CAPC) method. CAPC combines contrastive predictive coding with the Barlow Twins method, enhancing the model's ability to learn robust representations from unlabeled CSI time-series data. This approach improves model generalization, particularly when labeled data is scarce. CAPC also introduces a novel augmentation technique, dual view, which isolates free space propagation information from hardware distortions, further enhancing representation quality for WiFi sensing applications. Through extensive evaluations, this thesis demonstrates the effectiveness of both RSCNet and CAPC. RSCNet achieves results on par with the state-of-the-art performance in HAR tasks while drastically reducing computational burdens on edge devices. CAPC outperforms baseline SSL approaches and traditional supervised methods, showcasing its superior generalization capabilities in unseen environments. The dual view augmentation further enhances CAPC's performance by reducing electronic distortions. This thesis concludes that RSCNet and CAPC contribute significantly to the advancement of robust and practical wireless sensing technologies. These frameworks address critical challenges in the field, paving the way for wider adoption of WiFi sensing in real-world applications.
  • ItemOpen Access
    A Novel Vulnerable Smart Contracts Profiling Method Based on Advanced Genetic Algorithm Using Penalty Fitness Function
    (2024-11-07) HajiHosseinKhani, Sepideh; Habibi Lashkari, Arash
    With the advent of blockchain networks, there has been a transition from traditional contracts to Smart Contracts (SCs), which are crucial for maintaining trust within these networks. Previous methods for analyzing SC vulnerabilities typically lack accuracy and effectiveness, struggling to detect complex vulnerabilities due to limited data availability. This study introduces a novel approach to detecting and profiling SC vulnerabilities, featuring two components: a new analyzer named BCCC-SCsVulLyzer and an advanced Genetic Algorithm (GA) profiling method. The BCCC-SCsVulLyzer extracts 240 features, while the enhanced GA employs techniques such as Penalty Fitness Function and Adaptive Mutation Rate to profile vulnerabilities. Additionally, this work introduces a new dataset, BCCC-SCsVul-2024, with 111,897 Solidity source code samples for practical validation. Three taxonomies are established to enhance the efficiency of profiling techniques. Our approach demonstrated superior precision and accuracy, proving efficient in time and space complexity. The profiling technique also makes the model highly transparent and explainable, highlighting the potential of GA-based profiling to improve SC vulnerability detection and enhance blockchain security.
  • ItemOpen Access
    Exploring Mid-Air Gestures in Human-Computer Interfaces
    (2024-11-07) Fallah, Saba; MacKenzie, Scott
    This thesis investigates three hand gesture detection technologies and their integration into everyday tasks for interacting with technology, including text entry, target selection, and gaming. Through three user studies, these technologies have been incorporated and evaluated within novel applications relying on mid-air hand gestures. The first user study introduces a one-handed mid-air gesture-based text entry method in Virtual Reality with two layouts of a four-key ambiguous keyboard. Five participants took part in a five-day longitudinal study using a computer camera and an open-source gesture detection framework for gesture detection. Their entry speed (wpm) and error rate (%) were recorded and analyzed. In the second user study, a Leap Motion Controller (LMC) was integrated with a physical keyboard, introducing a novel computer keyboard, LeapBoard, that combines mid-air hand gestures with physical keys. The evaluation compared LeapBoard’s point-and-select ability to a touch-based method (using a touchpad) and a mid-air gesture-based method in a target selection task. A user study with 12 participants measured throughput (bps) and error rates (%) across different selection methods, movement amplitudes, and target widths. The third user study investigates the effects of two types of mid-air hand gesture-based in-put methods on children’s performance, fun, and preference and compares them to a mouse input method. The evaluation was done using a card-matching game on a laptop with 18 children between five and seven. The trial completion time (s), number of selected cards, and children’s perception regarding ease of use, likability, and willingness to play the game using each input method again were recorded and analyzed.
  • ItemOpen Access
    Examining the Effectiveness of Generative Artificial Intelligence for the Identification of Defeaters in Assurance Cases
    (2024-07-18) Khakzad Shahandashti, Kimya; Boaye Belle, Alvine
    Assurance cases are structured arguments that allow verifying the correct implementation of the created systems’ non-functional requirements (e.g., safety, security, reliability). This allows for preventing system failure. The latter may result in loss of life, severe injuries, large-scale environmental damage, property destruction, and major economic loss. Assurance cases support the certification of systems in compliance with industrial standards (e.g., DO-178C, ISO26262). However, the presence of assurance weakeners - deficits and logical fallacies - signals gaps in evidence and reasoning. Addressing this, our research presents a comprehensive taxonomy for categorizing these assurance weakeners, alongside proposed management strategies. The taxonomy divides weakeners into four categories of uncertainty: aleatory, epistemic, ontological, and argumentation. It also categorizes management approaches into representation, identification, and mitigation. A critical aspect of strengthening assurance cases involves identifying argumentation uncertainty or defeaters. To automate this process, we explore the capabilities of GPT-4 Turbo, a sophisticated large language model by OpenAI. We focus on its application in detecting defeaters within assurance cases represented using Eliminative Argumentation(EA) notation. Our initial evaluation assesses GPT-4 Turbo’s proficiency in understanding and applying this notation, a key factor in effectively generating defeaters. The results indicate that GPT-4 Turbo is highly adept in EA notation, demonstrating its potential to generate a diverse range of defeaters, thereby enhancing the robustness and reliability of assurance cases. Moreover, we used GPT-4 Turbo to identify defeaters which demonstrated effective proficiency.
  • ItemOpen Access
    Secure Abstraction of Fractionalization Smart Contracts for Non-Fungible Tokens
    (2024-07-18) Haouari, Wejdene; Fokaefs, Marios
    Non-fungible tokens (NFTs) have faced a downturn in interest, prompting a critical reassessment of their utility and accessibility. Fractionalization emerges as a solution, by enabling multiple parties to hold a stake in a single NFT, fractionalization lowers the barrier to entry for investors, enhancing market liquidity. Implementing fractionalization relies on smart contracts, which govern the terms of division, and transfer of fractions of an NFT. However, the complexity of these contracts and the immutable nature of blockchain underscores the importance of security. This thesis tackles the challenge of implementing fractionalization solutions and enhancing the security of supporting smart contracts. It thoroughly analyzes current fractionalization methods, identifies security vulnerabilities, and explores mitigation strategies to contribute to a safer and inclusive NFT ecosystem. The goal is to propose a baseline for standardizing NFT fractionalization to improve interoperability and address security concerns, laying the groundwork for a more unified and secure ecosystem.
  • ItemOpen Access
    Design and Automatic Generation of Safety Cases of ML-Enabled Autonomous Driving Systems
    (2024-07-18) Sivakumar, Mithila; Belle, Alvine Boaye
    Safety cases play a pivotal role in ensuring system reliability and acceptability, providing a structured argument supported by evidence. However, gaps in safety case literature hinder comprehensive safety assurance practices. In this thesis, we address this challenge through a three-fold approach. First, we conducted a bibliometric analysis following PRISMA 2020 guidelines to identify trends and knowledge gaps in safety assurance research. The analysis reveals critical areas lacking full safety cases and highlights the need for automated safety case construction. Then, we manually constructed a safety case for an ML-enabled component of an autonomous vehicle. Finally, leveraging large language models like GPT-4, we conducted experiments to automate safety case generation. Results indicate that GPT-4 produces safety cases with moderate accuracy and high semantic similarity to ground truth cases. This comprehensive methodology enhances safety practices, aiding researchers, analysts, and regulators in achieving robust safety assurance in complex systems.
  • ItemOpen Access
    Enhancing code review for improved code quality with language model-driven approaches
    (2024-03-16) Rahman, Shadikur; Prince, Enamul Hoque
    Code review is essential for maintaining software development standards, yet achieving effective reviews and issue resolution remains challenging. This thesis introduces RefineCode, an application tool to find actionable code reviews and provide similar code reviews as references within an organization, aiding developers in resolving issues effectively. To this end, we collected 9,500 code reviews from five private projects in an industrial setting and empirically evaluated various classification methods for identifying actionable code reviews. RefineCode automatically recommends relevant solutions from Stack Overflow based on textual similarity and entity linking between code reviews and Stack Overflow issues. Additionally, it integrates a chatbot feature, leveraging large language models to propose potential solutions for actionable code reviews. These features empower developers to make informed decisions, enhancing code quality by guiding issue resolution without reinforcing misunderstandings.
  • ItemOpen Access
    Towards Efficient and Robust Caching: Investigating Alternative Machine Learning Approaches for Edge Caching
    (2024-03-16) Torabi, Hoda ; Litoiu, Marin; Khazaei, Hamzeh
    This study introduces HR-Cache, a caching framework designed to enhance the efficiency of edge caching. The increasing complexity and variability of traffic classes at edge environments pose significant challenges for traditional caching methods, which often rely on simplistic metrics. HR-Cache addresses these challenges by implementing a learning-based strategy grounded in Hazard Rate ordering, a concept originally used to establish cache performance upper bounds. By employing a lightweight supervised machine learning model, HR-Cache learns from HR-based caching decisions and predicts the "cache-friendliness" of incoming requests, identifying "cache-averse" objects as priority candidates for eviction. Our experiment results demonstrate HR-Cache's superior performance. It consistently achieves 2.2–14.6% greater WAN traffic savings compared to the LRU strategy and outperforms both heuristic and state-of-the-art learning-based algorithms, while adding minimal prediction overhead. Though designed with the considerations of edge caching limitations, HR-Cache can be adapted with minimal changes for broader applicability in various caching contexts.
  • ItemOpen Access
    Trajectory Prediction Learning using Deep Generative Models
    (2024-03-16) Li, Jing; Papagelis, Manos
    Trajectory prediction involves estimating an object's future path using its current state and historical data, with applications in autonomous vehicles, robotics, and human motion analysis. Deep learning methods trained on historical data have been applied to this task, but they struggle with complex spatial dependencies due to the intricate nature of trajectory data and dynamic environments. We introduce TrajLearn, a novel trajectory prediction model using generative models and higher-order mobility flow representations (hexagons). TrajLearn, given a trajectory's recent history and current state, predicts its next k steps. It employs a variant of beam search for exploring multiple paths, ensuring spatial continuity. Our experiments demonstrate that TrajLearn surpasses current leading methods and other baselines by about 60% on various real-world datasets. We also explore different prediction horizons (k values), perform resolution sensitivity analysis, and conduct an ablation study to evaluate the contributions of different model components.
  • ItemOpen Access
    Multi-Versioning and Microservices: A Strategy for Developing Reliable Software Systems
    (2024-03-16) Akhtarian, Nazanin; Khazaei, Hamzeh
    In the dynamic realm of software engineering, adaptability is key to sustaining system performance and reliability. Software iterations often bring about challenges such as unexpected bugs and performance issues, necessitating a nuanced approach to maintain system integrity. In this work, we propose employing software multi-versioning to enhance system reliability. We embark on an in-depth exploration of the reliability of microservices within chaotic environments. Using Chaos Mesh, we simulate a series of disruptions in a microservices-based application, i.e., the Online Boutique. Through real experimentation, we systematically introduce various chaos disruptions, such as Pod failures, response delay, and memory stress, to investigate their impact on the system's reliability. We define a reliability metric that quantifies the robustness and efficiency of each software version under adverse conditions. Leveraging this metric, we introduce a dynamic controller that adjusts the population of each version, ensuring optimal resource distribution, reliability and system performance. Additionally, our research evaluates how the system adapts to varying workloads. We investigate how well the system can adjust its scalability—specifically, the number of replicas—in response to changes in \acrshort{cpu} usage as the user load fluctuates. Our findings demonstrates the system's capability to scale dynamically based on workload demands, ensuring robustness and efficiency. In conclusion, our study provides a detailed framework for employing software multi-versioning as a means to enhance system reliability. By devising a reliability metric and implementing a dynamic scaling system that responds to both reliability assessments and workload variations, we offer a comprehensive strategy to fortify systems against the unpredictable nature of software evolution, ensuring they remain resilient and make efficient use of resources.
  • ItemOpen Access
    Key-Frame Based Motion Representations for Pose Sequences
    (2024-03-16) Thasarathan, Harrish Patrick; Derpanis, Konstantinos
    Modelling human motion is critical for computer vision tasks that aim to perceive human behaviour. Extending current learning-based approaches to successfully model long-term motions remains a challenge. Recent works rely on autoregressive methods, in which motions are modelled sequentially. These methods tend to accumulate errors, and when applied to typical motion modelling tasks, are limited up to only four seconds. We present a non-autoregressive framework to represent motion sequences as a set of learned key-frames without explicit supervision. We explore continuous and discrete generative frameworks for this task and design a key-framing transformer architecture to distill a motion sequence into key-frames and their relative placements in time. We validate our learned key-frame placement approach with a naive uniform placement strategy and further compare key-frame distillation using our transformer architecture with an alternative common sequence modelling approach. We demonstrate the effectiveness of our method by reconstructing motions up to 12 seconds.