Publication:

Toward Generalizable Device Identification with Radio Frequency Fingerprint

 
cris.virtual.department#PLACEHOLDER_PARENT_METADATA_VALUE#
cris.virtual.department#PLACEHOLDER_PARENT_METADATA_VALUE#
cris.virtual.department#PLACEHOLDER_PARENT_METADATA_VALUE#
cris.virtual.department#PLACEHOLDER_PARENT_METADATA_VALUE#
cris.virtual.orcid0009-0003-3299-453X
cris.virtual.orcid0000-0003-1152-6617
cris.virtual.orcid0000-0003-0064-5020
cris.virtual.orcid0000-0002-1499-3782
cris.virtualsource.departmentc5e628ef-e419-4d62-9ba9-bea5b592e744
cris.virtualsource.departmenta2b34a52-0296-4181-842d-18e16639a1d7
cris.virtualsource.department0890472f-b07c-459b-b27e-54ab6db1557d
cris.virtualsource.departmentff3b6b19-1b9b-4d92-89ba-4b833fe2bef7
cris.virtualsource.orcidc5e628ef-e419-4d62-9ba9-bea5b592e744
cris.virtualsource.orcida2b34a52-0296-4181-842d-18e16639a1d7
cris.virtualsource.orcid0890472f-b07c-459b-b27e-54ab6db1557d
cris.virtualsource.orcidff3b6b19-1b9b-4d92-89ba-4b833fe2bef7
dc.contributor.authorNhem, Thayheng
dc.contributor.authorWeyn, Maarten
dc.contributor.authorPeeters, Michael
dc.contributor.authorBerkvens, Rafael
dc.date.accessioned2026-11-02T14:18:24Z
dc.date.available2026-11-02T14:18:24Z
dc.date.createdwos2026
dc.date.issued2026
dc.description.abstractRadio Frequency Fingerprint (RFF) is a widely investigated methodology for identifying wireless devices. It refers to unique signal characteristics caused by hardware imperfections in a device's radio front-end. While deep learning models have shown strong performance in classifying known devices using RFF, they struggle to generalize to unseen devices. It often produces overlapping and cluttered feature representations. In this work, we propose using the FlexiPrint framework with a contrastive center loss to extract features from any new devices and cluster them. We conduct experiments on a dataset of 25 devices transmitting Wi-Fi signals. We train different supervised models using 20 devices for comparison. We then test those models on the remaining 5 devices and show that our proposed approach can extract features from the unknown devices beyond the training set with a Silhouette clustering score of 0.6209, which is significantly higher than the baseline model with a score of 0.0441. We also visualize the learned features using t-SNE, which further confirms the improved clustering of device features, demonstrating the potential generalizability to new devices.
dc.description.wosFundingTextThis research is partially funded by Research Foundation - Flanders (FWO) PESSO project under grant number: G018522N.
dc.identifier.doi10.1109/wcnc65185.2026.11555343
dc.identifier.isbn979-8-3315-7730-8
dc.identifier.issn1525-3511
dc.identifier.urihttps://imec-publications.be/handle/20.500.12860/60680
dc.language.isoeng
dc.provenance.editstepusergreet.vanhoof@imec.be
dc.publisherIEEE
dc.relation.ispartofseriesIEEE Wireless Communications and Networking Conference
dc.source.beginpage1
dc.source.conferenceIEEE Wireless Communications and Networking Conference (WCNC)
dc.source.conferencedate2026-04-13
dc.source.conferencelocationKuala Lumpur
dc.source.endpage6
dc.source.journal2026 IEEE WIRELESS COMMUNICATIONS AND NETWORKING CONFERENCE, WCNC
dc.source.numberofpages6
dc.subject.keywordsOPEN-SET
dc.subject.keywordsTRANSMITTER
dc.title

Toward Generalizable Device Identification with Radio Frequency Fingerprint

dc.typeProceedings paper
dspace.entity.typePublication
imec.internal.crawledAt2026-06-17
imec.internal.sourcecrawler
imec.internal.wosCreatedAt2026-11-02
Files
Publication available in collections: