ETD

Archivio digitale delle tesi discusse presso l'Università di Pisa

Tesi etd-05302009-124852


Tipo di tesi
Tesi di laurea specialistica
Autore
CANDORE, ANDREA
Indirizzo email
andrea.candore@gmail.com, andrea.candore@alice.it
URN
etd-05302009-124852
Titolo
Robust Radiometric Fingerprinting for Wireless Devices
Dipartimento
INGEGNERIA
Corso di studi
INGEGNERIA DELLE TELECOMUNICAZIONI
Relatori
Relatore Prof.ssa Koushanfar, Farinaz
Relatore Dott. Garroppo, Rosario Giuseppe
Relatore Prof. Giordano, Stefano
Parole chiave
  • QPSK
  • Symbols
  • Phase
  • Identification
  • Classification
  • Magnitude
  • Metrics
  • Fingerprinting
Data inizio appello
22/06/2009
Consultabilità
Non consultabile
Data di rilascio
22/06/2049
Riassunto
Several previous works in wireless network security have demonstrated the weakness of conventional security mechanisms. Spoofing MAC address by transmission monitoring is not a big deal even for inexpert attackers and once a malicious device has obtained such an address the range of possible threats is wide.
In this thesis a new identification algorithm is proposed. This method is based on the unique and unclonable characteristics of radio transmitters. By exploiting minimal impairments inside analog hardware components a digital fingerprint can be extracted by only looking at the received signal and a database of legal clients can be built.
Unique characteristics of antenna, oscillator, amplifiers and other transmitter specific elements are indeed directly transferred on the transmitted signal providing idiosyncratic features.
For testing the model we use ten reconfigurable FPGA WARP radio boards as transmitters and a common WARP radio board as receiver. Signature extraction is done by looking at received frames and as much as 14 different classifiers are deployed. These metrics include frequency offset, modulated phase offset, in-phase/quadrature offset from the origin, and magnitude. Each classifier may provide a weak contribution, but it is only after combining all of them using a weighted voting algorithm that our model is able to perfectly classify six boards over ten and provide fair results for other three radios by analyzing as few as ten received frames per board.
File