IoT Telemetry Dataset and UNSW-NB15 Dataset
Description
IoT Telemetry Intrusion Detection Dataset: This dataset is a synthetically labelled IoT telemetry dataset derived from the original IoT telemetry data (iot_telemetry_data.csv). It contains 405,184 records and extends the original telemetry attributes with synthetic device identifiers, timestamps, binary attack labels, and attack categories for intrusion detection research. The dataset contains IoT sensor attributes including CO, humidity, light, LPG, motion, smoke, and temperature, together with device information. Based on the observed device and attack-type distributions in the original data, synthetic device_id, label, and attack_type attributes were generated to support controlled IDS experiments. The dataset contains 202,592 Normal records and 202,592 Attack records. The attack records are distributed across five categories: Data Injection, DoS, Malware, Replay, and Spoofing, resulting in six classes when Normal traffic is included. The synthetic dataset is intended for academic research, machine learning, deep learning, GAN-based intrusion detection, binary and multiclass classification, and reproducibility studies, and is also meant to assess the model's generalisability. It should be considered a synthetic, labelled extension of the original IoT telemetry dataset, not a replacement for the original data. The dataset can be used for developing and evaluating IoT intrusion detection models, including machine learning, deep learning, ensemble learning, and generative adversarial network (GAN)-based approaches. It supports binary and multiclass classification and evaluation using metrics such as accuracy, precision, recall, F1-score, false-positive rate, false-negative rate, and confusion matrices. The dataset is provided in CSV format and is intended for academic and research purposes. UNSW-NB15: The UNSW-NB15 dataset is a network intrusion detection dataset developed for evaluating machine learning and deep learning-based cybersecurity models. It contains normal network traffic as well as multiple categories of malicious activities, making it suitable for both binary and multiclass intrusion detection research. The dataset includes 49 features representing network flow, packet, protocol, and connection characteristics, along with the corresponding attack labels. The attack categories include Fuzzers, Analysis, Backdoors, DoS, Exploits, Generic, Reconnaissance, Shellcode, and Worms. UNSW-NB15 can be used to develop and evaluate Intrusion Detection Systems (IDS), particularly for network attack classification, anomaly detection, feature analysis, and machine learning model comparison. Common evaluation measures include accuracy, precision, recall, F1-score, false-positive rate, false-negative rate, and confusion matrices. The dataset is widely used in cybersecurity research and provides a challenging benchmark containing diverse modern network attack behaviours. It is provided in CSV format for research and experimental use.
Files
Steps to reproduce
Steps to Reproduce 1. Environment Setup Install Python 3.10 or later and create a virtual environment. Install the required packages: pip install numpy pandas scikit-learn matplotlib seaborn torch 2. IoT Telemetry Dataset Download the iot_telemetry_ids.csv file from the public repository. Load the dataset using Pandas. Check the 13 attributes, missing values, class labels, and class distribution. Use label for binary classification (Normal/Attack). Use attack_type for multiclass classification (Normal, Data Injection, DoS, Malware, Replay, and Spoofing). Separate input features from the target label. Encode categorical/device-related attributes and convert numerical attributes to an appropriate numeric format. Apply feature scaling where required by the selected model. Split the data into training and testing subsets using a fixed random seed for reproducibility. Train the selected IDS models using the same preprocessing and evaluation protocol. Evaluate the models using accuracy, precision, recall, F1-score, false positives, false negatives, and confusion matrices. 3. UNSW-NB15 Dataset Download the UNSW-NB15 dataset files from the public repository. Load the training and testing data using Pandas. Identify the 49 network traffic features and the corresponding attack labels. Preprocess categorical protocol/service/state attributes using appropriate encoding. Handle numerical features and apply scaling where required. Use the binary attack label for normal-versus-attack classification. For multiclass experiments, use the attack-category labels: Fuzzers, Analysis, Backdoors, DoS, Exploits, Generic, Reconnaissance, Shellcode, and Worms. Train the selected IDS models using the same experimental settings. Evaluate the results using accuracy, precision, recall, F1-score, false-positive rate, false-negative rate, and confusion matrices. 4. Reproducibility Use fixed random seeds, record the Python/library versions, retain the same train/test partitions, and report all model hyperparameters. Save the trained models, predictions, confusion matrices, and evaluation results to enable independent verification of the reported results.