IoT intrusion detection dataset for decentralized federated learning

The dataset comprises intrusion detection data collected from IoT crowdsensing devices (sourced from Electrosense) for the purpose of training decentralized federated learning models. It includes data from eight distinct IoT devices: Raspberry Pi 4 (devices 1 and 2) and Raspberry Pi 3B (devices 3 to 8). The anomalous data represent both normal operational behavior and behavior under infection by eight different types of malware, including Botnet, HttpBackdoor, Backdoor, TheTick, Beurk, Bdvl, XMRig Coinminer, and Ransomware. Device activity is recorded across multiple behavioral dimensions: system calls and processes, kernel events, file system operations, resource usage, input/output activities, and network behavior.The dataset is organized into two main components: the raw behavior logs, stored in the subdirectory 0_raw_collected_data, and the preprocessed feature representations, located in 1_syscall_processed_features. The total dataset size is 308.1 GB, comprising approximately 21 million behavioral records.