Datasets

The dataset collected over several years shows an example of natural drift in a real deployment. The images collected inherently exhibits non-stationary visual distributions. This dataset provides a realistic fine-grained benchmark for evaluating long-term model stability, drift-aware learning, and deployment-ready vision systems.

A dataset based of real-world sensor data, collected within an office environment at the University of Bristol. The office is actively used by a significant number of academic personnel and students (max occupancy of 28 people). It gets exposed to environmental changes such as seasonal temperature, humidity, and light fluctuations. The endpoints are located in different locations in the lab, showing locality and drift over time. Each IoT endpoint hosts sensors providing temperature, humidity, pressure, gas, accelerometer, and light readings. We collected two additional pieces of information, i.e., the measurements accuracy value, calculated by the environmental sensors, and the received signal strength indicator (RSSI).

The dataset consists of 315 captured walking sequences. Each sequence is simultaneously captured by two Azure Kinect devices. The two captures are interleaved to effectively double the frame rate. Fifteen participants partook in this experiment. Each experiment consists of seven walking actions, and having three predefined trajectories per experiment. That results in 21 sequences per participant. The data were collected using the Azure Kinect Sensor SDK. They were later processed using the official tools and libraries provided by Microsoft. For each sequence and trajectory, the positions and orientations of thirty-two tracked joints were obtained and saved.

Each streetlight is photographed by a Raspberry Pi Camera Module v1, installed on each lamppost, providing a unique camera placement, photographic angle, and distance from the streetlight. Several streetlights are partially obstructed by vegetation, outside the Field of View (FoV) of the Raspberry Pi camera or affected by conditions (e.g., rain, snow, direct sunlight, etc.). This is a unique and diverse dataset of images that can be used for training tools and machine learning models for inspection, monitoring and maintenance use-cases within Smart Cities applications.

A dataset based on the Linux Auditing System, which contains malicious and benign container activities. Two malicious scenarios were developed, i.e., a denial of service and a privilege escalation attack, where an adversary uses a container to compromise an edge device. The container activity is captured through the host system via system calls. The time series auditd dataset contains partial labels for the benign and malicious related system calls.

A large-scale dataset of ETSI ITS-G5 Dedicated Short Range Communications (DSRC) network interactions. Two vehicles, fitted with OBU transceivers, exchanged broadcasts Cooperative Awareness Messages (CAMs) with four RSUs, over the period of 8 recorded sessions of ∼2h each. The experiment was conducted over the licensed DSRC band, and the unlicensed ISM bands, in order to compare the behaviour and the affect of interference. Each transmitted and received CAM is logged along with its RSSI value and accurate positioning information. The data are presented in raw PCAP traces and post-processed CSV files.

Two vehicles, fitted with OBU transceivers, exchanged broadcasts Cooperative Awareness Messages (CAMs) with three RSUs, over the period of 6 recorded sessions of ∼2h each. The experiment was conducted over the licensed DSRC band. Each transmitted and received CAM is logged along with its RSSI value and accurate positioning information. The data are presented in raw PCAP traces and post-processed CSV files.