Pendekatan Convolutional Neural Network untuk Mengantisipasi Jamming-attack pada Wireless Sensor Network

Abstract

IEEE 802.11 wireless networks (Wi-Fi) deployed within Wireless Sensor Networks have become a cornerstone of modern communication infrastructures. However, the inherently open nature of the wireless transmission medium renders them susceptible to jamming attacks, which can markedly degrade quality of service or even precipitate complete communication failure. Consequently, the detection and mitigation of jamming attacks constitute a critical challenge for maintaining network reliability. This dissertation proposes a machine-learning approach for detecting jamming attacks in WLANs by employing a Convolutional Neural Network (CNN) that receives spectrogram images of the received signals as input. By converting time-domain signals into the frequency domain using the Short-Time Fourier Transform (STFT), a visual representation containing characteristic jamming features is produced; the CNN subsequently learns these features to achieve accurate classification. The CNN architecture is optimised experimentally to attain the best performance in discriminating between normal conditions and various forms of jamming. Beyond detection, the dissertation also introduces a mitigation mechanism via Adaptive Channel Switching (ACS) driven by the CNN’s classification output. When an attack is detected, the system automatically shifts the communication channel to one free from interference, based on adaptive logic, thereby preserving network connectivity in real-time. A Python-based simulation is conducted to evaluate the effectiveness of the combined CNN model and ACS strategy in enhancing the network’s resilience against attacks. The evaluation results indicate that the proposed approach attains a 96.40% classification accuracy in detecting jamming and significantly reduces the adverse impact of interference on network performance. The primary contribution of this research lies in the integration of spectrogram-based CNN detection with a dynamic, channel-switching mitigation strategy, which collectively bolsters the physical-layer security and resilience of IEEE 802.11 WLANs.