Examensarbeten för masterexamen

Browse

Senast publicerade

Visar 1 - 5 av 1764
  • Post
    Lightweight Data-Driven Anomaly Detection for IoT-Based Smart Grids: Capabilities and Limitations
    (2024) Eskilson, Lisa; Lager Carvalho, Alexander; Chalmers tekniska högskola / Institutionen för data och informationsteknik; Chalmers University of Technology / Department of Computer Science and Engineering; Duvignau, Romaric; Almgren, Magnus
    The integration of Internet of Things devices in critical infrastructure, such as the smart grid, has made it possible to monitor and manage energy distribution with increased efficiency. However, as these devices become more complex and interconnected, detecting physical tampering or data manipulation by malicious actors, such as the Sandworm attack in Ukraine, becomes increasingly challenging. One way that has shown promise in addressing this problem is the use of lightweight data-driven anomaly detection techniques. In this thesis, PASAD and USAD, two state-of-the-art lightweight data-driven anomaly detection algorithms were selected and evaluated using a series of experiments simulating common attacks against smart grids as suggested by relevant research. These experiments aimed to investigate the viability of these algorithms in IoT-based smart grids. More specifically, the experiments include two different attack areas, namely OT- and network-level attacks, which were crafted by manipulating real smart grid operational data. These experiments were evaluated using time series-aware metrics to get a fair assessment of the efficacy of the algorithms. The results from the experiments were used to evaluate the viability of lightweight data-driven anomaly detection algorithms and their capabilities and limitations were highlighted. Furthermore, the knowledge acquired from executing the experiments was used to propose guidelines for the development of an event management system that handles alerts produced by different models to provide valuable and actionable information to the OT operator. The selected algorithms were successful in detecting various long-duration attacks with stealth characteristics, while other, shorter and more direct attacks, were significantly harder to detect. Despite this, these lightweight data-driven anomaly detection algorithms proved to be a good fit for the experiments evaluated in this thesis.
  • Post
    Real-time Relevance: RAG with Dynamic Context for Improved Natural Language Responses
    (2024) Landgren, Malte; Giljegård, Oskar; Chalmers tekniska högskola / Institutionen för data och informationsteknik; Chalmers University of Technology / Department of Computer Science and Engineering; Johansson, Richard; Johansson, Richard
    Today’s Retrieval Augmented Generation (RAG) systems often struggle when trying to answer questions that require complex multi-hop reasoning. In this thesis we investigate an autoregressive Large Language Model (LLM) architecture which can generate a real-time relevant dense search vector for every token generation step. To facilitate this we also develop a synthetic data generation technique to acquire search query vector labels on a token-by-token level, requiring only a generating LLM and a document database. We investigate the quality of the synthetic data, and provide an attention based relabeling method which decreases hallucinations, improving the correctness of the labels by 67%. The architecture is able to produce query vectors 27 times faster than a separate embedder at the cost of retrieval accuracy. Finally, we train and employ the model in an active retrieval question-answering setting.
  • Post
    A Domain-Specific Language for Crossplatform, Edge-deployed Machine Learning Models: A Model Interpretation-based Approach
    Karlsson Landgren, Albin; Perhult Johnsen, Philip; Chalmers tekniska högskola / Institutionen för data och informationsteknik; Chalmers University of Technology / Department of Computer Science and Engineering; Heyn, Hans-Martin; Strüber, Daniel
    Deploying machine learning (ML) models on edge devices presents unique challenges. The challenges arise from the different environments used for developing ML models and those required for their deployment, leading to a gray area of competence and expertise between ML engineers and application developers. This thesis presents the design and implementation of a domain-specific language aimed at simplifying the deployment of ML models on edge devices, specifically smartphones. It aims to bridge the gap between ML engineers and application engineers, creating a shared platform for deploying ML models on edge devices. The study exists at the intersection of model-driven engineering, machine learning, and cross-platform smartphone development. It explores model-driven engineering in an environment where developers don’t have full control over the deployment platform, using model interpretation to generate ML serving pipelines (pre- and postprocessing of data before and after inference) during runtime, thus removing the need to re-release an application upon changes to a pipeline. We follow a design science approach consisting of three research cycles. We elicited requirements through an initial literature study and interviews with engineers at the collaboration company. This was followed by designing and implementing an artifact within the domain presented above. Finally, we evaluated the proposed solution with engineers at the collaboration company through a controlled experiment and subsequent qualitative interviews. The developed artifact consists of a lightweight, JSON-based domain-specific language designed to describe ML serving pipelines, along with an accompanying Flutter library to generate the pipelines during runtime. The evaluation showed that it increased development speed, decreased the amount of code required to make changes to an ML serving pipeline, and made engineers less experienced in mobile development more confident contributing to the domain.
  • Post
    Learning to Navigate Over Stochastic Transport Networks Using Multi-Armed Bandits: A Contextual Approach for Efficient Online Learning in Road Network Graphs with Multi-Armed Bandits to Minimize Long- Term Travel Time
    (2024) Nilsson, Hannes; Johansson, Rikard; Chalmers tekniska högskola / Institutionen för data och informationsteknik; Chalmers University of Technology / Department of Computer Science and Engineering; Dubhashi, Devdatt; Haghir Chehreghani, Morteza
    As part of the ongoing process of phasing out fossil fuel vehicles, attempts have been made to extend the effective range and adoption rate of electric vehicles through navigation systems focused on energy consumption. One way to approach this problem is by viewing route selections as a multi-armed bandit problem. This allows the system to adapt and recommend better routes over time, to minimize energy consumption. For navigation systems to be useful in practice, guiding vehicles from one point to another in minimal time is crucial. Therefore, this project examines the effectiveness of multi-armed bandit algorithms for time-efficient navigation in complex real-world environments, without initial information. For this purpose, we adapt a previously studied online learning framework developed for energy efficiency, and extract road segment travel time distributions from the traffic simulation software SUMO. The framework is applied to the Luxembourg road network and our results demonstrate that contextual multi-armed bandits using tree ensembles are highly effective. More specifically, we show that TEUCB and TETS, which we implement using both XGBoost and random forest, outperform state-of-the-art contextual multi-armed bandits based on neural networks and linear models. Further, by additional comparison of TEUCB and TETS to other bandit algorithms based on tree models, we identify at least two properties to explain their high level of performance. First, tree ensemble methods appear to offer relatively accurate travel time predictions from the contextual information available in this problem. Second, the ability to generalize over different arms and infer the travel time on one road segment from observations gathered on other ones, based on similarities, seems highly advantageous for this problem.
  • Post
    Evaluating Machine Learning Algorithms in Design Pattern Recognition - Exploring the Performance of Classification and Clustering Algorithms in Design Pattern Recognition Utilising Large Language Models
    (2024) Andersson, Simon; Berggren, Viktor; Chalmers tekniska högskola / Institutionen för data och informationsteknik; Chalmers University of Technology / Department of Computer Science and Engineering; Heyn, Hans-Martin; Horkoff, Jennifer
    Design Pattern Recognition (DPR) is an ongoing research challenge in the field of software engineering for increasing software maintainability in code. Recent work has utilised Large Language Models (LLMs) for extracting semantic information from code. This study follows up on previous research and investigates, explores, and evaluates the performance of multiple classification and clustering algorithms when applied to embeddings extracted from LLMs. Performance is explored between contexts using different LLMs, design patterns, and programming languages. Data for design pattern implementations was gathered for Java, Python, and C# via GitHub and the P-MARt repository. Each algorithm was run with tuned hyperparameters, and their average performance across multiple runs was compared. The results indicate variance for the individual performance of the algorithms, but the overall performance order between the algorithms remains the same. Classification algorithms outperformed clustering algorithms, and clustering algorithms had low performance in the measured metrics across all tests. The results also showed a difference in performance between behavioral, creational, and structural design patterns. This study shows further promise for the use of LLMs for DPR and recognises the need for larger studies utilising LLMs for DPR.