https://www.mdu.se/

Machine learning is increasingly called upon to guide decisionsin critical industrial applications. Its predictive powerpromises gains in efficiency, yet its black-box nature and lackof guarantees pose risks in contexts where behavior must remainanalyzable and safe. This thesis asks how machine learning can bemade trustworthy, explainable, and efficient enough for engineersto deploy in practice. Three gaps hamper broader adoption. Few works provide formal orstatistical guarantees on ML outputs paired with explanationsthat engineers can act on (Gap~A). Data-driven models thatgeneralize across hardware configurations without retrainingremain rare, and existing simulators are prohibitively slow(Gap~B). Many contributions address individual components ofindustrially motivated problems without combining them intovalidated end-to-end pipelines (Gap~C). To address Gap~A, we apply abstraction to neural networks,showing that inputs with negligible effect on the output can beformally identified and removed, producing simpler yet boundedmodels open to verification. We then introduce a conformalprediction framework for CPU load forecasting that providesstatistically guaranteed coverage intervals, combined withShapley value analysis to trace individual task contributions tothe predicted load. To address Gap~B, we develop a data-drivencache memory surrogate using long short-term memory networks,reproducing cache miss distributions across unseen hardwareconfigurations at a fraction of the simulator's computationalcost. To address Gap~C, we present HASCO, a Hybrid AI SimulationCompiler that translates natural language accident reports intoexecutable vehicular simulation scenarios through a structuredcompilation approach with deterministic validation. Together, these contributions establish a path toward machinelearning that is not merely powerful but trustworthy, explainable,and practically deployable in the industrial workflow.

The rapid growth of the Internet of Things (IoT) has led to increasingly complex and data-intensive systems, often with conflicting or interdependent application requirements. Many IoT applications have strict timing requirements, particularly for safety-critical services, e.g., in the industrial domain. However, due to the vastness of the IoT ecosystem, decisions about where, how, and what timing attributes to measure are often ad hoc and context-specific, fragmenting the community’s understanding of what “timing” entails. Moreover, security and privacy requirements are often overlooked within IoT. Common approaches to uphold confidentiality, integrity, and availability are often not suitable for resource-constrained IoT devices, as they may introduce unacceptable overhead. Consequently, designing a secure and time-critical IoT system is a challenging task. 

These challenges motivate the goal of this licentiate thesis, which is to support timing and security in IoT systems through edge-centric approaches. To achieve this, the research follows an Action Design Research-inspired process and combines systematic literature reviews, surveys, and experimental evaluations using real IoT devices. The findings include a consolidated view of timing definitions, a cross-domain synthesis of explicitly reported timing-related requirements, and a unified categorization of timing metrics in the literature. In addition, comparative experiments of edge-centric data-reduction techniques highlight accuracy and reduction trade-offs for different applications, while a study of lightweight security mechanisms demonstrates how protocol choices and configuration influence timing behavior. Overall, this thesis contributes foundational knowledge for reasoning about timing and security in IoT systems by establishing key definitions, requirements, and metrics, and by evaluating selected technical mechanisms to meet these requirements.