BPBiLSTM-IDS: a lightweight intrusion detection framework for cyber-physical UAV networks

Description

Hafiz Muhammad Attaullah and seven colleagues built in 2026 a lightweight intrusion detector for drone networks, where wireless links invite serious attacks. Drones now fly autonomous missions, but their networks face denial-of-service, false-data injection, replay, and fake access-point attacks that corrupt data and disrupt control. Standard detectors tend to raise too many false alarms and demand too much computing for small flying platforms with shifting topologies.

Their framework first selects the most useful traffic features with binary pigeon optimisation, a search method inspired by flocking birds. It then detects attacks with a bidirectional long short-term memory network paired with a one-dimensional convolutional network. On a cyber-physical drone dataset the combination reached 98.74 percent accuracy across ten runs, beating traditional approaches, with high precision and recall and a low false-positive rate. The authors present it as a scalable, adaptive guard for real-time drone security.

Catalog record

Source: https://europepmc.org/articles/PMC13473474