"FIRE-C2: Environment-Triggered Backdoor Attacks in Wireless IoT Networks"

"We present the FIRE-C2 dataset, a comprehensive multi-modal dataset designed for studying environmentally triggered command-and-control (C2) backdoor attacks in IoT-based cyber-physical systems. Generated using the ns-3 network simulator, the dataset models a smart-building fire monitoring network comprising 80 wireless sensor nodes arranged in an 8\u00d710 grid topology, with 5 compromised attacker nodes (6.25% penetration rate). The dataset comprises 249 successful simulation runs, totaling approximately 18.3 hours of simulated network activity, which yields over 2.7 million packet-level records, 2.6 million node-state records, and 111,443 ground-truth covert channel events. Six synchronized CSV files capture complementary data modalities: (i) packet-level network communications with timing, routing, and wireless metrics; (ii) node-state time-series including physical sensor readings, temperature dynamics, and security status; (iii) covert channel activities documenting timing-modulated beacons and LSB-encoded exfiltration; (iv) network performance metrics; (v) attack lifecycle events; and (vi) fire propagation dynamics. Traffic composition includes 70.1% fire-related telemetry, 17.4% benign background traffic, 9.6% C2 backdoor traffic, and 2.9% dropped packets. All records share common keys (timestamp, node_id, run_id), enabling cross-modal analysis. The dataset supports four benchmark intrusion detection tasks of varying difficulty: node-level sensor-rich detection, node-level network-only detection, time-windowed detection, and packet-level classification."