Details

Autor(en) / Beteiligte

• Iglesias, Cristovao F.
• Guo, Rongchen
• Nucci, Pedro
• Miceli, Claudio
• Bolic, Miodrag

Titel

Automated Extraction of IoT Critical Objects from IoT Storylines, Requirements and User Stories via NLP

Ist Teil von

• 2023 10th IEEE Swiss Conference on Data Science (SDS), 2023, p.104-107

Ort / Verlag

IEEE

Erscheinungsjahr

2023

Quelle

IEEE Xplore

Beschreibungen/Notizen

• The first step to designing a resilient Internet of Things (IoT) application is to identify IoT critical objects (services, devices and resources) in the design phase. However, this step is a time-intensive task, because they are manually identified from storylines, requirements and user stories and have other challenges. In this work, we assessed the usefulness of Named Entity Recognition (NER) models to automatically identify IoT critical objects as a way to make a modelling process faster and less prone to errors. This was performed with the development of five NER models based on five different architectures (Spacy, BERT, Transformers, LSTM-CRF and ELMo) that were trained and tested with a large dataset with 7396 annotated sentences. Our results indicate that all NER models had satisfactory performance, but BERT had the best one and can be useful to support the time-intensive step of the early stages of the development of resilient IoT systems. Furthermore, these NER models have a high potential to be extended to a framework to automatically extract IoT critical objects from documents (storyline and requirements) and list all possible IoT threats and resilient countermeasures that can be used in the design of a resilient IoT application.