Search for: All records

Time series analysis has evolved from traditional autoregressive models to deep learning, Transformers, and Large Language Models (LLMs). While vision models have also been explored along the way, their contributions are less recognized due to the predominance of sequence modeling. However, challenges such as the mismatch between continuous time series and LLMs’ discrete token space, and the difficulty in capturing multivariate correlations, have led to growing interest in Large Vision Models (LVMs) and Vision-Language Models (VLMs). This survey highlights the advantages of vision models over LLMs in time series analysis, offering a comprehensive dual-view taxonomy that answers key research questions like how to encode time series as images and how to model imaged time series. Additionally, we address pre- and post-processing challenges in this framework and outline future directions for advancing the field. 

To study the security properties of the Internet of Things (IoT), firmware analysis is crucial. In the past, many works have been focused on analyzing Linux-based firmware. Less known is the security landscape of MCU-based IoT devices, an essential portion of the IoT ecosystem. Existing works on MCU firmware analysis either leverage the companion mobile apps to infer the security properties of the firmware (thus unable to collect low-level properties) or rely on small-scale firmware datasets collected in ad-hoc ways (thus cannot be generalized). To fill this gap, we create a large dataset of MCU firmware for real IoT devices. Our approach statically analyzes how MCU firmware is distributed and then captures the firmware. To reliably recognize the firmware, we develop a firmware signature database, which can match the footprints left in the firmware compilation and packing process. In total, we obtained 8,432 confirmed firmware images (3,692 unique) covering at least 11 chip vendors across 7 known architectures and 2 proprietary architectures. We also conducted a series of static analyses to assess the security properties of this dataset. The result reveals three disconcerting facts: 1) the lack of firmware protection, 2) the existence of N-day vulnerabilities, and 3) the rare adoption of security mitigation.