Past Textual content-to-SQL for IoT Protection: A Complete Framework for Querying and ClassifyingIoT Threats
Authors: Ryan Pavlich, Nima Ebadi, Richard Tarbell, Billy Linares, Adrian Tan, Rachael Humphreys, Jayanta Kumar Das, Rambod Ghandiparsi, Hannah Haley, Jerris George, Rocky Slavin, Kim-Kwang Raymond Choo, Glenn Dietrich, Anthony Rios
Summary: Recognizing the promise of pure language interfaces to databases, prior research have emphasised the event of text-to-SQL techniques. Whereas substantial progress has been made on this subject, present analysis has targeting producing SQL statements from textual content queries. The broader problem, nevertheless, lies in inferring new details about the returned information. Our analysis makes two main contributions to handle this hole. First, we introduce a novel Web-of-Issues (IoT) text-to-SQL dataset comprising 10,985 text-SQL pairs and 239,398 rows of community visitors exercise. The dataset comprises extra question varieties restricted in prior text-to-SQL datasets, notably temporal-related queries. Our dataset is sourced from a wise constructing’s IoT ecosystem exploring sensor learn and community visitors information. Second, our dataset permits two-stage processing, the place the returned information (community visitors) from a generated SQL may be categorized as malicious or not. Our outcomes present that joint coaching to question and infer details about the info can enhance general text-to-SQL efficiency, practically matching considerably bigger fashions. We additionally present that present giant language fashions (e.g., GPT3.5) wrestle to deduce new details about returned information, thus our dataset offers a novel take a look at mattress for integrating complicated domain-specific reasoning into LLM