Not logged in.
Quick Search - Contribution
Contribution Details
| Type | Bachelor's Thesis |
| Scope | Discipline-based scholarship |
| Title | Implementation and Detection of Spectrum Sensing Data Falsification Attacks Affecting Crowdsensing Platforms |
| Organization Unit | |
| Authors |
|
| Supervisors |
|
| Language |
|
| Institution | University of Zurich |
| Faculty | Faculty of Business, Economics and Informatics |
| Date | 2022 |
| Abstract Text | The usage of mobile data has increased massively over the past few years and the trend is only rising. A part of this trend is due to the growth of the Internet-of-Things (IoT), which is merging the digital and physical worlds. IoT devices are collecting and transmitting countless of bits over the wireless spectrum and as a result, the radio frequency (RF) spectrum is getting bursty and overcrowded. Yet IoT devices are also beneficial for the RF spectrum, as they are used as sensors in monitoring networks that analyze the spectrum usage to optimize the use of the wireless spectrum. However, these devices are well-known to be resource-constrained and therefore a growing cybersecurity concern. In a sensing network, they are vulnerable to Spectrum Sensing Data Falsification (SSDF) attacks trying to manipulate the data. Recent research has proposed behavioral fingerprinting and Machine/Deep Learning (ML/DL) to detect those attacks. To improve the limitations of the recent literature, another implementation of the latest defined SSDF attacks is proposed in this thesis. The sensing software used in the crowdsensing monitoring platform ElectroSense has been modified to implement seven SSDF attacks. The attacks have been executed in several different configurations whilest the behavior of the infected device has been observed based on the system call trace. A Machine Learning (ML) framework thereafter has cleaned the gathered datasets, extracted features and trained multiple unsupervised ML algorithms with normal behavior data. The infected data has then been classified by the models to evaluate the anomaly detection performance in different settings. The experiments have demonstrated that the proposed implementation using variables is not reliably detectable compared to previous implementations using files stored in disk. |
| PDF File |
Download
|
| Export |
BibTeX
|