A novel RF fingerprinting approach for hardware ... - Semantic Scholar
j o u r n a l o f i n f o r m a t i o n s e c u r i t y a n d a p p l i c a t i o n s 1 9 ( 2 0 1 4 ) 1 4 3 e1 4 8
Available online at www.sciencedirect.com
ScienceDirect journal homepage: www.elsevier.com/locate/jisa
A novel RF fingerprinting approach for hardware integrated security Mohamed Kheir*, Heinz Kreft, Reinhard Kno¨chel Institute of Electrical and Information Engineering, Christian-Albrechts-University of Kiel, Kiel, Germany
article info
abstract
Article history:
A novel approach for RF fingerprinting using a simple microwave passive device is pro-
Available online 19 March 2014
posed. This device is a multi-port cavity resonator filled with a dielectric material mixed with randomly-distributed micro- or nano-particles. Such structure generates a typical
Keywords:
Physical Unclonable Function (PUF) that can be perfectly used for storing and protecting
Physical Unclonable Function (PUF)
secure information. This mixture guarantees a spatial random distortion on the electro-
RF fingerprint
magnetic fields and consequently a unique fingerprint. The extraction of these fingerprints
RF identification (RFID)
is based on the scattered near-fields which are represented by the S-parameters. Pre-
Scattering parameters
liminary measurement results show a high degree of reliability and reproducibility of the fingerprints over the UWB frequency range. ª 2014 Elsevier Ltd. All rights reserved.
1.
Introduction
Many emerging technologies require highly resilient, cost efficient and operational information security systems. Today’s security systems typically exist in form of a Subscriber Identity Module (SIM), Smart Card or as Trusted Platform Module (TPM). All these systems are based, more or less, on the inability to clone or to extract the stored secrets. However, such security assumption is proven to be invalid and the physical security seems to vanish (Kreft and Adi, 2012). As a possible wayout of this dilemma, Physical Unclonable Functions (PUFs) were proposed in the last decade to create physically unique entities as a physical security anchor required for many applications. This paper is aimed at developing a novel RF fingerprinting technology which is capable of producing and using these PUFs as a new cryptographic-primitive (Kreft and Adi, 2012; Kreft, 2012; Kheir et al., Oct. 2013; Kheir et al., Sep. 2013). This approach is based on advanced micro and nano materials
in combination with microwave technology. Some recent attempts have investigated potential solutions for data security and counterfeiting problems (Lakafosis et al., 2011; Preradovic and Karmakar, 2010; Pursula et al., 2011). However, most of the traditional RFID and fingerprinting approaches are mainly based on far-field antenna Tx/Rx and not an integrated authentication methodology. The proposed system does not include any antennas and does not require any transmission process that may introduce additional sources of errors and can be insecure. It is rather based on directly-coupled nearfield probes. The novel idea in this paper lies in using a cavity structure (the Cocoon-PUF) that can be further developed to fit in the common credit-card shape or as a standard chip housing. The proposed structure is intended to be a “proof-ofconcept” used for experimental and investigation purposes. It is realized for testing different mixtures and particles before the chip integration process since the final fingerprinting system will be fully embedded for on-chip integrated security purposes. In the following sections, the cavity structure as well as the measured S-parameters are going to be introduced.
* Corresponding author. E-mail addresses: [email protected] (M. Kheir), [email protected] (H. Kreft), [email protected] (R. Kno¨chel). http://dx.doi.org/10.1016/j.jisa.2014.02.001 2214-2126/ª 2014 Elsevier Ltd. All rights reserved.
144
2.
j o u r n a l o f i n f o r m a t i o n s e c u r i t y a n d a p p l i c a t i o n s 1 9 ( 2 0 1 4 ) 1 4 3 e1 4 8
Prototype design
The proposed microwave passive structure that generates the PUF is a four-port cylindrical cavity resonator. The cavity is made of Brass alloy (60% Zinc, 40% Copper) to maintain lower losses and meanwhile better environmental immunity. It has a total diameter of 95 mm including the screw fixtures with 25 mm height. This cavity is filled with a dielectric material (Polystyrene) with a dielectric constant of 3.3 and a loss tangent of 0.002. This dielectric filling is the fixing matrix mixed with micro-structured metallic particles. The material of these particles is an alloy of eight substances with a majority of Aluminum, Copper and Zinc with specific mixing percentages. The main feature of this Cocoon-PUF structure lies in the purely-randomized distribution of particles within the polystyrene filling. This dielectric filling is mixed with the metallic particles to provide a means for fixed and unclonable function due to the EM near-fields interacting with this mixture. This unclonable function (fingerprint) is impossible to regenerate due to the three-dimensional complexity of this mixture. Any attempt to intrude the device will automatically result in a totally changed fingerprint and demolish the key. The cylindrical cavity is excited by four coaxial SMA
connectors that are asymmetrically aligned in order to have different independent transmission paths. It is also intended to eliminate any effect due to the geometrical similarity of the port alignment. This structure was implemented in the central workshop, University of Kiel, utilizing some advanced production technologies after liquefying the dielectric before the distribution process. Six different samples are implemented for investigation purposes. Fig. 1 shows a photograph of four different samples used in this experiment. The first sample (#01), which is also identical to samples #02 and #06, is the only purelydielectric filled cavity with no particle inclusion. These undoped prototypes are used for comparison purposes. The other three samples, #03, #04 and #05, are filled with different particle masses as declared in Table 1. The amount of distributed particles per sample ranges from few hundreds (for the big-sized particles) up to millions (for the small-sized particles). Each particle has a spherical shape where the radius varies from 100 up to 900 mm. This illustrated “one-way function” transforms the irreversible fingerprint data into a reproducible key string applying noise cancellation and error correction. A minimum key space size of 256 independent bits is expected to obtain after this randomness extraction.
3.
Measurement results
A series of measurements is carried out in this section so as to verify the operation and functionality of the fabricated samples. These measurements are performed over the frequency range (1e15 GHz) using an Agilent E8361A PNA vector network analyzer. The maximum number of sweep points (20,000 points) is chosen for these measurements in order to obtain the highest possible accuracy over the whole frequency span. ShorteOpeneLoadeThrough (SOLT) calibration procedure is performed using an Agilent coaxial calibration kit. For an nport scattering matrix, there are nðn 1Þ=2 different transmission combinations which correspond to six independent fingerprints.
3.1.
Scattering parameters
The magnitudes of two different port-pair transmission parameters, namely S21 and S34, are shown in Fig. 2 for two zeroparticle samples with the serial numbers #01 and #02 as a first check for the measurement reliability and reproducibility. It can be observed from the same figure that both samples show almost identical scattering properties over the whole frequency span. One common feature of both samples is the resonance frequency that appears at 1.78 GHz. This frequency typically corresponds to the TM010 mode for the given cavity dimensions and filling material. This feature can be considered as a part of a unique signature for each sample since this
Fig. 1 e A photograph of the fabricated Cocoon-PUF prototypes. (a) A view of the open cavity structure. (b) A top-view shows the port connections.
Table 1 e Mass of metallic particles in all prototypes. Sample#
01
02
03
04
05
06
Mass of particles (g)
e
e
130
130
15
e
j o u r n a l o f i n f o r m a t i o n s e c u r i t y a n d a p p l i c a t i o n s 1 9 ( 2 0 1 4 ) 1 4 3 e1 4 8
145
Fig. 2 e Measured S-parameters of the two samples (#01 and #02).
resonance is strongly controlled by the filling parameters of each cavity. As will be proven shortly, the effective medium characteristics are influenced by the amount of mixed micro/ nano particles and consequently the effective permittivity and conductivity will be changed. Fig. 3 depicts another important measurement that is carried out to investigate the aging effect on the fabricated prototype #03. The shown S21 and S34 curves combine the results of the same sample. However, the first measurement data (Old) has been recorded four months prior to the new (Recent) one. Actually, some shifts between both curves in addition to some changes in the characteristic points can be observed. This happened due to the fact that the two
measurements “Old” and “Recent” have been carried out at two different room temperatures. Plus the dielectric material (Polystyrene) has a sensitive nature to temperature which made these changes. Another two samples filled with the same material and particle mass and volume but with another random distribution manner. These two samples, namely #03 and #04, enclose 130 g of particle-mixture with three different radii. In Fig. 4, it is depicted that both “identical” samples have different scattering characteristics and consequently different fingerprints. Two more fully-different samples with different particles mass, size and distribution are shown in Fig. 5 where big
Fig. 3 e Measured S-parameters of the sample #02 at two diverse points of time.
146
j o u r n a l o f i n f o r m a t i o n s e c u r i t y a n d a p p l i c a t i o n s 1 9 ( 2 0 1 4 ) 1 4 3 e1 4 8
Fig. 4 e Measured S-parameters of the two samples (#03 and #04).
differences can be clearly observed that proves the validity of this approach. Another interesting phenomenon can also be observed that the fundamental resonance frequency of sample #05 is greatly deviated from the other sample (1.28 GHz). The interpretation of this phenomenon lies in the fact that sample #03 encloses much more particles compared to sample #05. Such big amount of metallic particles has significantly affected the effective permittivity/permeability of the medium which then corresponded to a change in the resonance frequency.
3.2.
Similarity measures
distinguish between different fingerprints. One possible method for assessing the similarity/dissimilarity between a pair of data-sets is evaluating the well-known Euclidean distance metric dE. This method is adopted in this section to accurately judge how similar or dissimilar the measured fingerprints are in terms of the calculated distance metrics (Theodoridis and Koutroumbas, 2006). In the same time, it can be an initial indicator for the uniqueness of the produced and reproduced fingerprints. The distance between two variables a and b with k-dimensions is calculatedvas ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi u k uX 2 (1) aj bj dE ¼ t j¼1
Similarity between any pair of the fabricated prototypes is an essential measure that is needed to compare and
Fig. 6 depicts the calculated Euclidean metrics in dB
Fig. 5 e Measured S-parameters of the two samples (#03 and #05).
j o u r n a l o f i n f o r m a t i o n s e c u r i t y a n d a p p l i c a t i o n s 1 9 ( 2 0 1 4 ) 1 4 3 e1 4 8
147
Fig. 6 e Euclidean distance metrics between different pairs of samples.
between each sample-pairs {(#03, #04), (Old, Recent), (#01, #02), (#01, #06)} referred to S21. The highest distance clearly appears between the two zero-particle samples and the filled ones. The other two curves, which represent the zero-particle samples, are quite close and the distance between them does not exceed 8 dB all over the displayed frequency range. However, the distance between both red and blue curves and the other green curve is higher than 20 dB which declares a higher dissimilarity between the samples (#03 and #04) themselves and the other pairs. As for the green curve, which compares samples #03 and #04, the distance is above 25 dB for most frequency points all over the band. As a rule of thumb, smaller distance between the fingerprints corresponds to a higher similarity. A more accurate decision on the similarity or uniqueness of each fingerprint has to be given by the aid of a specific “Fuzzy Extractor”. All distance calculations give an initial insight about the generated functions before implementing the embedded system together with its fuzzy extractor.
4.
Conclusion
A new technology for RF fingerprinting and security has been introduced. The proposed concept has shown reliable, reproducible and unclonable fingerprints that can be efficiently utilized for different RFID applications. Six different proof-ofconcept samples have been implemented and characterized in the frequency range (1e15) GHz in order to investigate the validity of the concept. Similarity metrics have been evaluated using the Euclidean distance which have verified the uniqueness of fingerprints. It has also been proven that all measurements are reproducible and the extracted fingerprints strongly depend on the amount, geometry and
distribution manner of the mixed particles. A deeper investigation on the temperature and environmental dependencies is planned as an extended study of this work.
Acknowledgments The authors wish to thank Mr. Matthias Burmeister of the central workshop of University of Kiel and his team for their considerable efforts in manufacturing the prototypes.
references
Kreft H, Adi W. Cocoon-PUF, a novel mechatronic secure element technology. In: 2012 NASA/ESA Conference on Adaptive Hardware and Systems (AHS); June 2012. pp. 227e32. Kreft H. Tamper-protected hardware and methods for using same. International Patent Application PCT/EP2012/054248, European Patent Office file: WO 2012/123400 A1, March 2012. Kheir M, Kreft H, Kno¨chel R. Nanostructured RFID technology: attack-response and temperature-robustness investigations. In: 43rd European Microwave Conference (EuMC), Nuremberg e Germany; Oct. 2013. Kheir M, Kreft H, Kno¨chel R. RF identification and security technique for highly-secure UWB information systems. In: 24th IEEE International Symposium on Personal, Indoor and Mobile Radio Communications (PIMRC), London e UK; Sep. 2013. Lakafosis V, Traille A, Lee Hoseon, Gebara E, Tentzeris MM, DeJean GR, et al. RF fingerprinting physical objects for anticounterfeiting applications. IEEE Trans Microw Theory Tech Feb. 2011;59(2):504e14. Preradovic S, Karmakar NC. Design of chipless RFID tag for operation on flexible laminates. IEEE Antenn Wireless Propag Lett 2010;9:207e10.
148
j o u r n a l o f i n f o r m a t i o n s e c u r i t y a n d a p p l i c a t i o n s 1 9 ( 2 0 1 4 ) 1 4 3 e1 4 8
Pursula P, Donzelli F, Seppa H. Passive RFID at millimeter waves. IEEE Trans Microw Theory Tech Aug. 2011;59(8):2151e7.
Theodoridis S, Koutroumbas K. Pattern recognition. 3rd ed. USA: Elsevier; 2006 (Chapter. 11).
Available online at www.sciencedirect.com
ScienceDirect journal homepage: www.elsevier.com/locate/jisa
A novel RF fingerprinting approach for hardware integrated security Mohamed Kheir*, Heinz Kreft, Reinhard Kno¨chel Institute of Electrical and Information Engineering, Christian-Albrechts-University of Kiel, Kiel, Germany
article info
abstract
Article history:
A novel approach for RF fingerprinting using a simple microwave passive device is pro-
Available online 19 March 2014
posed. This device is a multi-port cavity resonator filled with a dielectric material mixed with randomly-distributed micro- or nano-particles. Such structure generates a typical
Keywords:
Physical Unclonable Function (PUF) that can be perfectly used for storing and protecting
Physical Unclonable Function (PUF)
secure information. This mixture guarantees a spatial random distortion on the electro-
RF fingerprint
magnetic fields and consequently a unique fingerprint. The extraction of these fingerprints
RF identification (RFID)
is based on the scattered near-fields which are represented by the S-parameters. Pre-
Scattering parameters
liminary measurement results show a high degree of reliability and reproducibility of the fingerprints over the UWB frequency range. ª 2014 Elsevier Ltd. All rights reserved.
1.
Introduction
Many emerging technologies require highly resilient, cost efficient and operational information security systems. Today’s security systems typically exist in form of a Subscriber Identity Module (SIM), Smart Card or as Trusted Platform Module (TPM). All these systems are based, more or less, on the inability to clone or to extract the stored secrets. However, such security assumption is proven to be invalid and the physical security seems to vanish (Kreft and Adi, 2012). As a possible wayout of this dilemma, Physical Unclonable Functions (PUFs) were proposed in the last decade to create physically unique entities as a physical security anchor required for many applications. This paper is aimed at developing a novel RF fingerprinting technology which is capable of producing and using these PUFs as a new cryptographic-primitive (Kreft and Adi, 2012; Kreft, 2012; Kheir et al., Oct. 2013; Kheir et al., Sep. 2013). This approach is based on advanced micro and nano materials
in combination with microwave technology. Some recent attempts have investigated potential solutions for data security and counterfeiting problems (Lakafosis et al., 2011; Preradovic and Karmakar, 2010; Pursula et al., 2011). However, most of the traditional RFID and fingerprinting approaches are mainly based on far-field antenna Tx/Rx and not an integrated authentication methodology. The proposed system does not include any antennas and does not require any transmission process that may introduce additional sources of errors and can be insecure. It is rather based on directly-coupled nearfield probes. The novel idea in this paper lies in using a cavity structure (the Cocoon-PUF) that can be further developed to fit in the common credit-card shape or as a standard chip housing. The proposed structure is intended to be a “proof-ofconcept” used for experimental and investigation purposes. It is realized for testing different mixtures and particles before the chip integration process since the final fingerprinting system will be fully embedded for on-chip integrated security purposes. In the following sections, the cavity structure as well as the measured S-parameters are going to be introduced.
* Corresponding author. E-mail addresses: [email protected] (M. Kheir), [email protected] (H. Kreft), [email protected] (R. Kno¨chel). http://dx.doi.org/10.1016/j.jisa.2014.02.001 2214-2126/ª 2014 Elsevier Ltd. All rights reserved.
144
2.
j o u r n a l o f i n f o r m a t i o n s e c u r i t y a n d a p p l i c a t i o n s 1 9 ( 2 0 1 4 ) 1 4 3 e1 4 8
Prototype design
The proposed microwave passive structure that generates the PUF is a four-port cylindrical cavity resonator. The cavity is made of Brass alloy (60% Zinc, 40% Copper) to maintain lower losses and meanwhile better environmental immunity. It has a total diameter of 95 mm including the screw fixtures with 25 mm height. This cavity is filled with a dielectric material (Polystyrene) with a dielectric constant of 3.3 and a loss tangent of 0.002. This dielectric filling is the fixing matrix mixed with micro-structured metallic particles. The material of these particles is an alloy of eight substances with a majority of Aluminum, Copper and Zinc with specific mixing percentages. The main feature of this Cocoon-PUF structure lies in the purely-randomized distribution of particles within the polystyrene filling. This dielectric filling is mixed with the metallic particles to provide a means for fixed and unclonable function due to the EM near-fields interacting with this mixture. This unclonable function (fingerprint) is impossible to regenerate due to the three-dimensional complexity of this mixture. Any attempt to intrude the device will automatically result in a totally changed fingerprint and demolish the key. The cylindrical cavity is excited by four coaxial SMA
connectors that are asymmetrically aligned in order to have different independent transmission paths. It is also intended to eliminate any effect due to the geometrical similarity of the port alignment. This structure was implemented in the central workshop, University of Kiel, utilizing some advanced production technologies after liquefying the dielectric before the distribution process. Six different samples are implemented for investigation purposes. Fig. 1 shows a photograph of four different samples used in this experiment. The first sample (#01), which is also identical to samples #02 and #06, is the only purelydielectric filled cavity with no particle inclusion. These undoped prototypes are used for comparison purposes. The other three samples, #03, #04 and #05, are filled with different particle masses as declared in Table 1. The amount of distributed particles per sample ranges from few hundreds (for the big-sized particles) up to millions (for the small-sized particles). Each particle has a spherical shape where the radius varies from 100 up to 900 mm. This illustrated “one-way function” transforms the irreversible fingerprint data into a reproducible key string applying noise cancellation and error correction. A minimum key space size of 256 independent bits is expected to obtain after this randomness extraction.
3.
Measurement results
A series of measurements is carried out in this section so as to verify the operation and functionality of the fabricated samples. These measurements are performed over the frequency range (1e15 GHz) using an Agilent E8361A PNA vector network analyzer. The maximum number of sweep points (20,000 points) is chosen for these measurements in order to obtain the highest possible accuracy over the whole frequency span. ShorteOpeneLoadeThrough (SOLT) calibration procedure is performed using an Agilent coaxial calibration kit. For an nport scattering matrix, there are nðn 1Þ=2 different transmission combinations which correspond to six independent fingerprints.
3.1.
Scattering parameters
The magnitudes of two different port-pair transmission parameters, namely S21 and S34, are shown in Fig. 2 for two zeroparticle samples with the serial numbers #01 and #02 as a first check for the measurement reliability and reproducibility. It can be observed from the same figure that both samples show almost identical scattering properties over the whole frequency span. One common feature of both samples is the resonance frequency that appears at 1.78 GHz. This frequency typically corresponds to the TM010 mode for the given cavity dimensions and filling material. This feature can be considered as a part of a unique signature for each sample since this
Fig. 1 e A photograph of the fabricated Cocoon-PUF prototypes. (a) A view of the open cavity structure. (b) A top-view shows the port connections.
Table 1 e Mass of metallic particles in all prototypes. Sample#
01
02
03
04
05
06
Mass of particles (g)
e
e
130
130
15
e
j o u r n a l o f i n f o r m a t i o n s e c u r i t y a n d a p p l i c a t i o n s 1 9 ( 2 0 1 4 ) 1 4 3 e1 4 8
145
Fig. 2 e Measured S-parameters of the two samples (#01 and #02).
resonance is strongly controlled by the filling parameters of each cavity. As will be proven shortly, the effective medium characteristics are influenced by the amount of mixed micro/ nano particles and consequently the effective permittivity and conductivity will be changed. Fig. 3 depicts another important measurement that is carried out to investigate the aging effect on the fabricated prototype #03. The shown S21 and S34 curves combine the results of the same sample. However, the first measurement data (Old) has been recorded four months prior to the new (Recent) one. Actually, some shifts between both curves in addition to some changes in the characteristic points can be observed. This happened due to the fact that the two
measurements “Old” and “Recent” have been carried out at two different room temperatures. Plus the dielectric material (Polystyrene) has a sensitive nature to temperature which made these changes. Another two samples filled with the same material and particle mass and volume but with another random distribution manner. These two samples, namely #03 and #04, enclose 130 g of particle-mixture with three different radii. In Fig. 4, it is depicted that both “identical” samples have different scattering characteristics and consequently different fingerprints. Two more fully-different samples with different particles mass, size and distribution are shown in Fig. 5 where big
Fig. 3 e Measured S-parameters of the sample #02 at two diverse points of time.
146
j o u r n a l o f i n f o r m a t i o n s e c u r i t y a n d a p p l i c a t i o n s 1 9 ( 2 0 1 4 ) 1 4 3 e1 4 8
Fig. 4 e Measured S-parameters of the two samples (#03 and #04).
differences can be clearly observed that proves the validity of this approach. Another interesting phenomenon can also be observed that the fundamental resonance frequency of sample #05 is greatly deviated from the other sample (1.28 GHz). The interpretation of this phenomenon lies in the fact that sample #03 encloses much more particles compared to sample #05. Such big amount of metallic particles has significantly affected the effective permittivity/permeability of the medium which then corresponded to a change in the resonance frequency.
3.2.
Similarity measures
distinguish between different fingerprints. One possible method for assessing the similarity/dissimilarity between a pair of data-sets is evaluating the well-known Euclidean distance metric dE. This method is adopted in this section to accurately judge how similar or dissimilar the measured fingerprints are in terms of the calculated distance metrics (Theodoridis and Koutroumbas, 2006). In the same time, it can be an initial indicator for the uniqueness of the produced and reproduced fingerprints. The distance between two variables a and b with k-dimensions is calculatedvas ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi u k uX 2 (1) aj bj dE ¼ t j¼1
Similarity between any pair of the fabricated prototypes is an essential measure that is needed to compare and
Fig. 6 depicts the calculated Euclidean metrics in dB
Fig. 5 e Measured S-parameters of the two samples (#03 and #05).
j o u r n a l o f i n f o r m a t i o n s e c u r i t y a n d a p p l i c a t i o n s 1 9 ( 2 0 1 4 ) 1 4 3 e1 4 8
147
Fig. 6 e Euclidean distance metrics between different pairs of samples.
between each sample-pairs {(#03, #04), (Old, Recent), (#01, #02), (#01, #06)} referred to S21. The highest distance clearly appears between the two zero-particle samples and the filled ones. The other two curves, which represent the zero-particle samples, are quite close and the distance between them does not exceed 8 dB all over the displayed frequency range. However, the distance between both red and blue curves and the other green curve is higher than 20 dB which declares a higher dissimilarity between the samples (#03 and #04) themselves and the other pairs. As for the green curve, which compares samples #03 and #04, the distance is above 25 dB for most frequency points all over the band. As a rule of thumb, smaller distance between the fingerprints corresponds to a higher similarity. A more accurate decision on the similarity or uniqueness of each fingerprint has to be given by the aid of a specific “Fuzzy Extractor”. All distance calculations give an initial insight about the generated functions before implementing the embedded system together with its fuzzy extractor.
4.
Conclusion
A new technology for RF fingerprinting and security has been introduced. The proposed concept has shown reliable, reproducible and unclonable fingerprints that can be efficiently utilized for different RFID applications. Six different proof-ofconcept samples have been implemented and characterized in the frequency range (1e15) GHz in order to investigate the validity of the concept. Similarity metrics have been evaluated using the Euclidean distance which have verified the uniqueness of fingerprints. It has also been proven that all measurements are reproducible and the extracted fingerprints strongly depend on the amount, geometry and
distribution manner of the mixed particles. A deeper investigation on the temperature and environmental dependencies is planned as an extended study of this work.
Acknowledgments The authors wish to thank Mr. Matthias Burmeister of the central workshop of University of Kiel and his team for their considerable efforts in manufacturing the prototypes.
references
Kreft H, Adi W. Cocoon-PUF, a novel mechatronic secure element technology. In: 2012 NASA/ESA Conference on Adaptive Hardware and Systems (AHS); June 2012. pp. 227e32. Kreft H. Tamper-protected hardware and methods for using same. International Patent Application PCT/EP2012/054248, European Patent Office file: WO 2012/123400 A1, March 2012. Kheir M, Kreft H, Kno¨chel R. Nanostructured RFID technology: attack-response and temperature-robustness investigations. In: 43rd European Microwave Conference (EuMC), Nuremberg e Germany; Oct. 2013. Kheir M, Kreft H, Kno¨chel R. RF identification and security technique for highly-secure UWB information systems. In: 24th IEEE International Symposium on Personal, Indoor and Mobile Radio Communications (PIMRC), London e UK; Sep. 2013. Lakafosis V, Traille A, Lee Hoseon, Gebara E, Tentzeris MM, DeJean GR, et al. RF fingerprinting physical objects for anticounterfeiting applications. IEEE Trans Microw Theory Tech Feb. 2011;59(2):504e14. Preradovic S, Karmakar NC. Design of chipless RFID tag for operation on flexible laminates. IEEE Antenn Wireless Propag Lett 2010;9:207e10.
148
j o u r n a l o f i n f o r m a t i o n s e c u r i t y a n d a p p l i c a t i o n s 1 9 ( 2 0 1 4 ) 1 4 3 e1 4 8
Pursula P, Donzelli F, Seppa H. Passive RFID at millimeter waves. IEEE Trans Microw Theory Tech Aug. 2011;59(8):2151e7.
Theodoridis S, Koutroumbas K. Pattern recognition. 3rd ed. USA: Elsevier; 2006 (Chapter. 11).
