Authentication system and device including physical unclonable function and threshold cryptography

Abstract

A device comprising: a physical unclonable function (PUF) device configured to generate an output value based on hardware characteristics of the PUF device; and a processor connected to the PUF device, the processor configured to: execute a cryptographic operation in a sequence of ordered stages including a first stage and a second stage, the executing comprising: in the first stage: recovering a first secret value based on a first output value obtained from the PUF device; executing a first sub-operation using the first secret value; and removing unobscured values from memory prior to execution of a subsequent stage; in the second stage: recovering a second secret value based on a second output value obtained from the PUF device; and executing a second sub-operation using the second secret value to enable execution of a cryptographic operation encoded with at least the first secret value and the second secret value.

Description

FIELD OF THE INVENTION

This disclosure relates generally to hardware verification, and in particular but not exclusively, to binding authentication to protect against tampering and subversion by substitution.

BACKGROUND OF THE INVENTION

The unique properties of PUFs provide several advantages to cryptographic constructions. In general, PUFs may provide some or all of three main advantages: (1) eliminating private key storage, (2) providing tamper detection, and (3) establishing a hardware root-of-trust. Private key storage can be eliminated by evaluating a PUF to dynamically regenerate a value unique to an identified piece of hardware having that PUF. As to tamper detection, a PUF's unclonable properties (e.g., wire delays, resistance) may be such that modification to the PUF irreversibly alters the PUF's mapping from challenges (inputs) to responses (outputs) after enrollment (however, not against malicious modifications before enrollment, e.g., Becker et al., “Stealthy Dopant-Level Hardware Trojans,” Cryptographic Hardware and Embedded Systems—CHES 2013, volume 8086 of Lecture Notes in Computer Science, pages 197-214, Springer, 2013). These PUF properties may be used to produce a hardware-unique, tamper-protected value from which a hardware root-of-trust can be established.

Literature on physical unclonable functions evaluates the properties of PUF hardware design (e.g., Gassend et al., “Silicon Physical Random Functions,” Proceedings of the 9th ACM conference on Computer and communications security, CCS '02, pages 148-160, ACM, 2002; Katzenbeisser et al., “PUFs: Myth, Fact or Busted A Security Evaluation of Physically Unclonable Functions (PUFs) Cast in Silicon,” Cryptographic Hardware and Embedded Systems—CHES '12, pages 283-301, Springer, 2012; Ravikanth, Physical one-way functions, Ph.D. thesis, 2001; Rührmair et al., “Applications of High-Capacity Crossbar Memories in Cryptography,” IEEE Trans. Nanotechnol., volume 10, no. 3:489-498, 2011; Suh et al., “Physical Unclonable Functions for Device Authentication and Secret Key Generation,” Proceedings of the 44 th annual Design Automation Conference, DAC '07, pages 9-14, ACM, 2007; Yu et al., “Recombination of Physical Unclonable Functions,” GOMACTech, 2010), provides formal theoretical models of PUF properties, and designs protocols around those definitions (cf. Armknecht et al., “A Formalization of the Security Features of Physical Functions,” Proceedings of the 2011 IEEE Symposium on Security and Privacy, SP '11, pages 397-412, IEEE Computer Society, 2011; Brzuska et, ah, “Physically Uncloneable Functions in the Universal Composition Framework,” Advances in Cryptology—CRYPTO 2011—31st Annual Cryptology Conference, volume 6841 of Lecture Notes in Computer Science, page 51, Springer, 2011; Frikken et al., “Robust Authentication using Physically Unclonable Functions,” Information Security, volume 5735 of Lecture Notes in Computer Science, pages 262-277, Springer, 2009; Handschuh et al., “Hardware Intrinsic Security from Physically Unclonable Functions,” Towards Hardware-Intrinsic Security, Information Security and Cryptography, pages 39-53, Springer, 2010; Kirkpatrick et al., “PUF ROKs: A Hardware Approach to Read-Once Keys,” Proceedings of the 6th ACM Symposium on Information, Computer and Communications Security, ASIACCS '11, pages 155-164, ACM, 2011; Parai et al., “Reliable and Efficient PUF-based Key Generation using Pattern Matching,” IEEE International Symposium on Hardware-Oriented Security and Trust (HOST), pages 128 -133, 2011; Rührmair et al., “PUFs in Security Protocols: Attack Models and Security Evaluations,” 2013 IEEE Symposium on Security and Privacy, volume 0:286-300, 2013; van Dijk et al., “Physical Unclonable Functions in Cryptographic Protocols: Security Proofs and Impossibility Results,” Cryptology ePrint Archive, Report 2012/228, 2012; Wu et al., “On Foundation and Construction of Physical Unclonable Functions,” 2010; Yu et al., “Lightweight and Secure PUF Key Storage using Limits of Machine Learning,” Proceedings of the 13th international conference on Cryptographic Hardware and Embedded Systems, CHES'11, pages 358-373, Springer, 2011).

Prior art PUF-based protocols fall into two broad categories: (1) a simple challenge-response provisioning process like the one described below in Protocol 3, or (2) cryptographic augmentation of a device's PUF response such that the raw PUF output never leaves the device. These approaches may require external entities to handle auxiliary information (e.g., challenges and their associated helper data) that is unsupported or superfluous in existing public key cryptography standards, and/or involve a hardware device authenticating to a challenge applied during an initial enrollment process, and/or are premised on the hardware device always recovering essentially the same response to a given challenge.

While a given challenge-response pair reflects the hardware state of a device when the pair was collected, the device will age and its hardware state drift over time. As the PUF hardware ages, the number of errors present in the responses may increase. Maiti et al. (“The Impact of Aging on an FPGA-Based Physical Unclonable Function,” International Conference on Field Programmable Logic and Applications (FPL), pages 151-156, 2011) study the effects of simulated aging on PUF hardware by purposefully stressing the devices beyond normal operating conditions. By varying both temperature and voltage, the authors were able to show a drift in the intra-PUF variation that, over time, will lead to false negatives. Maiti et al. note that the error drift strictly affected the intra-PUF error rate distribution tending towards the maximum entropy rate of 50%. After enough time elapses, the hardware device may no longer be able to recover the proper response for the enrolled challenge.

For example, assume that a specific challenge c_i is issued to a device during enrollment, with the device returning a public token {commitment_i, helper_i} that links the device's hardware identity with the challenge c. To be authenticated, the device uses the pair {c_i, helper_i} to recover its private identity _i^priv. As shown in FIG. 10, over time the PUF hardware may reach a time (e.g., at time τ=5 in the example of FIG. 10, which for simplicity assumes a drift that occurs linearly over time) at which hardware aging has increased the errors beyond the device's error correction limit, and the device is no longer able to reliably regenerate its private key.

Kirkpatrick et al. (“Software Techniques to Combat Drift in PUF-based Authentication Systems,” Workshop on Secure Component and System Identification, 2010) describe a method for detecting hardware aging drift, and responding by updating the device's challenge-commitment pair stored on an external server. This approach requires that the server maintain auxiliary information in the form of challenge-commitment pairs, however, and that a periodic protocol be executed between the server and the device.

Another challenge facing PUF-based systems is side channel attacks, which seek to observe and analyze auxiliary environmental variables to deduce information about the sensitive PUF output. For example, electromagnetic (EM) analysis (e.g., Merli et al., “Semi-invasive EM Attack on FPGA RO PUFs and Countermeasures,” Proceedings of the Workshop on Embedded Systems Security, WESS '11, pages 2:1-2:9, ACM, 2011; Merli et al., “Side-Channel Analysis of PUFs and Fuzzy Extractors,” Trust and Trustworthy Computing, volume 6740 of Lecture Notes in Computer Science, pages 33-47, Springer, 2011; Schuster, Side-Channel Analysis of Physical Unclonable Functions (PUFs), Master's thesis, Technische Universitat. Munchen, 2010) extracts PUF output bits by observing changing EM fields during device operation. Another side channel attack methodology is (simple or differential) power analysis (e.g., Karakoyunlu et al., “Differential template attacks on PUF enabled cryptographic devices,” IEEE International Workshop on Information Forensics and Security (WIFS), pages 1-6, 2010; Kocher et al., “Introduction to Differential Power Analysis,” Cryptography Research, Inc., 2011; Kocher et al., “Differential Power Analysis,” Proceedings of the 19th Annual International Cryptology Conference on Advances in Cryptology, CRYPTO '99, pages 388-397, Springer, 1999; Rührmair et al., “Power and Timing Side Channels for PUFs and their Efficient Exploitation,” 2013), where power traces are collected from a device and analyzed to extract sensitive information (e.g., PUF output bits). Over many observations of a device recovering essentially the same response to a fixed challenge, an adversary can discover the sensitive PUF output.

While it is known that the effectiveness of side channel attacks may in some systems be reduced by introducing randomness (Coron, “Resistance Against Differential Power Analysis For Elliptic Curve Cryptosystems,” Cryptographic Hardware and Embedded Systems, volume 1717 of Lecture Notes in Computer Science, pages 292-302, Springer, 1999), disguising sensitive values in this way may leave some vulnerability since the underlying values remain static and/or introduce additional complexity and/or processing overhead.

SUMMARY OF THE INVENTION

In an authentication system according to the present invention, a PUF's challenge-response behavior may be internalized and used to maintain shares of a private key. This approach may be implemented such that a PUF-enabled hardware device can execute arbitrary threshold cryptographic operations (e.g., decryption, digital signature generation, zero knowledge proofs) without ever generating, reconstructing, or storing a private key. It may also be implemented so as to eliminate the need for any external entity to issue challenges and store helper data for a device, and/or so as to enable PUF-based protocols that an external entity cannot distinguish from standard public key protocols. In one embodiment, a device may be equipped with a root-of-trust such as a PUF and configured to periodically refresh all sensitive values that must be generated, recovered, or processed by the device. This may be employed to mitigate PUF aging and/or side-channel attacks. The threshold share operations may be staggered such that one share always remains stored.

BRIEF DESCRIPTION OF THE DRAWINGS

FIG. 1 is a functional diagram of a device having a single PUF circuit and two threshold shares;

FIG. 2 is a functional diagram of a device having dual PUF circuits;

FIG. 3 is a graph illustrating an attempt to interpolate an example polynomial P(x) of degree d=3 given only one point, which fails as at least d+1 points are needed:

FIG. 4 illustrates a failed attempt to interpolate the same polynomial P(x) given two points;

FIG. 5 illustrates a failed attempt to interpolate the same polynomial P(x) given three points;

FIG. 6 illustrates a successful interpolation of P(x) given four points;

FIG. 7 is an operational flowchart of enrollment of a device like that of FIG. 2 in an embodiment of the invention:

FIG. 8 is an operational flowchart of threshold cryptographic operations in a device like that of FIG. 2 in an embodiment of the invention;

FIG. 9 is an operational flowchart of the staggered threshold operations;

FIG. 10 is a graph illustrating errors over time in PUF output for a fixed challenge;

FIG. 11 is a graph illustrating errors over time in PUF output with an updated challenge-helper pair for τ=1;

FIG. 12 is a graph illustrating errors over time in PUF output, with pair refreshing establishing a bounded PUF error rate;

FIG. 13 is a diagram depicting a batch PUF processing center; and

FIG. 14 is a diagram depicting joint identity from composite elements.

DETAILED DESCRIPTION OF EMBODIMENTS

The present invention is described with reference to the example of an embodiment utilizing elliptic curve cryptography (including the associated terminology and conventions), but the inventive concept and teachings herein apply equally to various other cryptographic schemes such as ones employing different problems like discrete logarithm or factoring (in which regard the teachings of U.S. Pat. No. 8,918,647 are incorporated here by reference), and the invention is not limited by the various additional features described herein that may be employed with or by virtue of the invention. Before setting forth details of the invention, basic handling of PUF output, modeling assumptions, and primitives for PUF-based cryptographic schemes and threshold cryptography applicable to the example embodiment are described.

PUF output is noisy in that it varies slightly despite evaluating the same input. This is generally addressed with fuzzy extraction, a method developed to eliminate noise in biometric measurements. (See Juels et al., “A Fuzzy Commitment Scheme,” Proceedings of the 6th ACM conference on Computer and Communications Security, CCS '99, pages 28-36, ACM, 1999). Fuzzy extraction may in part be employed within a device having a PUF such as within an auxiliary control unit, such that the output is constant for a fixed input. Fuzzy extraction (or reverse fuzzy extraction) may for example employ a “secure sketch,” as described by Juels et al.

A secure sketch SS for input string O, where ECC is a binary (n, k, 2t+1) error correcting code of length n capable of correcting t errors and V←{0,1}^k is a k-bit value, may be defined as SS(O; V)=O⊕ECC(V). This definition can be used to build a Gen algorithm, which outputs a set V, P, where V is the value to be reconstructed and P is a helper string (which may be public) that is used to recover V.

Algorithm 1 Gen Algorithm
Input : A finite field   n of order n; A challenge c
O ← PUF(c)
V ∈  n
P ← O ⊕ ECC(V)
return   V, P

Correspondingly, a Rep algorithm can be defined such that, on input O′ within a maximum Hamming distance t of O, the original value V may be recovered. Rep(O′, P), where D is the decoding scheme for the binary (n, k, 2t+1) error-correcting code ECC and O′ is an input such that dist(O,O′)≤t, can be defined as:

$\begin{array}{c}\mathrm{Rep}\left(O^{\prime },P\right)=D\left(P\oplus O^{\prime }\right)\\ =D\left(O\oplus \mathrm{ECC}\left(V\right)\oplus O^{\prime }\right)\\ =V\end{array}$

This definition can then be used to build a Rep algorithm that allows a PUF output O′ that differs from the original output O by at most t to reproduce output V such that Rep(O′)=V using the helper string P=O⊕ECC(V):

Algorithm 2 Rep Algorithm
Input : A challenge c, Helper string P
O′ ← PUF(c)
V ← D(P ⊕ O′)
return V

Gen and Rep algorithms such as these may be used in PUF-based protocols to ensure that the same value V is recovered so long as the PUF outputs O, O′ differ by at most t bits.

It is desirable that an adversary cannot predict a device's PUF response r for a challenge c with more than negligible probability (at least without physical access to the device), and that helper data does not reveal anything to an adversary about PUF responses. In assessing these security aspects, the following entities may be considered: a set of servers , where each server s_i∈ authentication of devices on its system; a set of devices d_i∈, each with an embedded PUF; and an adversary that wishes to masquerade as a legitimate device d_i∈ to obtain resources stored on some subset of the servers ′⊆. It may be assumed that all entities are bound to probabilistic polynomial-time (PPT), i.e., can only perform computation requiring polynomially many operations with respect to a global security parameter λ (which refers to the number of bits in the relevant parameter). Computation requiring exponentially many operations with respect to λ is not efficient for the agents, and will succeed with only negligible probability.

Games can be employed to assess a PPT adversary's advantage in (1) predicting a PUF's output, and (2) distinguishing helper data from a truly random string. It is sufficient to require that an adversary's advantage in a game is negligible with respect to the security parameter of the protocol, where a function ƒ(x): is negligible if for every positive polynomial (⋅) and sufficiently large x, the absolute value of ƒ(x) is less than 1/(x). Although we describe exemplary games for capturing properties (1) and (2), they may be substituted for other game formulations designed to capture the concepts.

The unpredictability of a PUF can be assessed through the following game between an adversary and a PUF device P: {0,1}^κ1>{0,1}^κ2 mapping input strings from the challenge space _P p of length κ₁ to the response space _P of length κ₂ where λ is the security parameter for the protocol, given in unary as 1^λ.

Protocol 1: PUF - PRED: PUF Prediction Game
		Adversary		PUF Device P
	(1)	ci ∈  P ⊂  P,	→
		0 ≤ i ≤ poly(λ)
			←	ri = P(ci) ∈  P	(2)
				P ⊂  P,
				0 ≤ i ≤ poly(λ)
	(3)	Challenge c      P	→
	(4)	c′i ∈    P′ ⊂  P,	→
		c        P′,
		0 ≤ i ≤ poly(λ)
			←	r′i = P(ci′) ∈  P′	(5)
				P′ ⊂  P,
				0 ≤ i ≤ poly(λ)
	(6)	Guess r’     P(c)	→

The game proceeds as follows:

1. The adversary issues polynomially many (w.r.t. the security parameter λ) challenges c_i∈_P the PUF device P, where the challenge set _P is a proper subset of the entire challenge space _P.

2. The PUF device P returns the responses {r_i|r_i←P(c_i)} to .

3. The adversary eventually outputs a challenge c that was not in the original set of challenge queries _P. The adversary is not allowed to query the PUF device P on the committed challenge c.

4. The adversary may once again issue a new set of polynomially many challenges c′_i∈′_P to the PUF device P. The adversary is not allowed to query the PUF device P on the committed challenge c.

5. The PUF device P returns the responses {r′_i|r′_i←P(c′_i)} to .

6. The adversary eventually outputs a guess r′ for P's response to the committed challenge c.

The probability of the adversary guessing the correct response of the PUF P to the challenge c is then ^-PREDκ₂):=Pr[r=r′]. That is, the adversary only wins the game when 's guess r′ is equal to P's actual response r←P(c) to 's committed challenge c. (Technically, the PUF's output is noisy and will vary slightly on any fixed input; thus, the equality is taken with respect to the output of a fuzzy extractor (e.g., Dodis et al.)).

In the PUF indistinguishability game, an adversary is asked to differentiate between the output r of the fuzzy extractor for a PUF P and a randomly chosen string s∈ of the same length .

Protocol 2: PUF - IND: PUF Indistinguishability Game
	Adversary		PUP Device P
(1)	ci ∈       ⊂  P,	→	(Ri, Hi) ←
	0 ≤ i ≤ poly(λ)		Gen(ri = P(c))
		←	Hi ∈  P ⊂  P,	(2)
			0 ≤ i ≤ poly(λ)
(3)	ci ∈     ⊂  P,	→
	0 ≤ i ≤ poly(λ)
		←	Ri ∈  P ⊂  P,	(4)
			0 ≤ i ≤ poly(λ)
(5)	Challenge c	→	b ∈ {0, 1}
		←	b(s ∈ {0, 1    +	(6)
			(1 − b)(Ri),
			Ri = Rep(P(c), Hi)
(7)	c′i ∈       ⊂  P,	→
	c ≠ c′i,
	0 ≤ i ≤ poly(λ)
		←	R′i ∈    P ⊂  P,	(8)
			0 ≤ i ≤ poly(λ)
(9)	Guess b'     b	→

This game proceeds as follows:

1. Adversary executes the enrollment phase on any challenge c_i∈_P.

2. The PUF device returns the corresponding helper string H_i from the output of Gen. Denote this set of challenge-helper pairs (c_i, H_i) as .

3. Adversary now requests the PUF response r_i=P(c_i) for any c_i∈. Denote the set of requested challenges in this step .

4. For all requests c_i∈, the PUF device returns the set {r_i|r_i←P(c_i)}.

5. Adversary selects a challenge c∉, such that has H_i but not R_i for c. The PUF device chooses a bit b∈{0,1} uniformly at random.

6. If b=0, is given R_i=Rep(P(c)=r_i, H_i). Otherwise, if b=1 then is given a random string s∈{0,1}^l.

7. Adversary is allowed to query the PUF device for c′_i∈ so long as no c′_i=c.

8. For all requests c′_i≠, the PUF device returns the set {r′_i|r′_i←P(c′_i)}.

9. The adversary outputs a guess bit b′, and succeeds when b′=b.

Types of PUFs

Rührmair et al. (“Modeling Attacks on Physical Unclonable Functions,” Proceedings of the 17th ACM conference on Computer and communications security, CCS '10, pages 237-249, ACM, 2010) define three distinct classes of PUF devices:

1. A Weak PUF is typically used only to derive a secret key. The challenge space may be limited, and the response space is assumed to never be revealed. Typical constructions include the SRAM (Holcomb et al., “Initial SRAM State as a Fingerprint and Source of True Random Numbers for RFID Tags,” In Proceedings of the Conference on RFID Security, 2007), Butterfly (Kumar et al., “Extended abstract: The Butterfly PUF Protecting IP on Every FPGA,” IEEE International Workshop on Hardware-Oriented Security and Trust, pages 67-70, 2008), Arbiter (Lee et al., “A technique to build a secret key in integrated circuits for identification and authentication applications,” IEEE Symposium on VLSI Circuits: Digest of Technical Papers, pages 176-179, 2004), Ring Oscillator (Suh et al.), and Coating (Tuyls et al., “Read-Proof Hardware from Protective Coatings,” Proceedings of the 8th international conference on Cryptographic Hardware and Embedded Systems, CHES'06, pages 369-383, Springer, 2006) PUFs.

2. A Strong PUF is assumed to be (i) physically impossible to clone, (if) impossible to collect a complete set of challenge response pairs in a reasonable time (typically taken to be on the order of weeks), and (iii) difficult to predict the response to a random challenge. For example, the super-high information content (SHIC) PUF described by Rührmair et al. (“Applications of High-Capacity Crossbar Memories in Cryptography”) may be considered a Strong PUF.

3. A Controlled PUF satisfies all of the criteria for strong PUFs, and additionally implements an auxiliary control unit capable of computing more advanced functionalities to cryptographically augment protocols. A controlled PUF should ideally satisfy the five conditions below.

One definition for an ideal physical unclonable function P_d: {0,1}^κ1→{0,1}^κ2 bound to a device d that the function preferably exhibits the following properties:

1. Unclonable: Ideally, Pr[dist(y,x)≤t|x←U_κ1, y←P(x), z←P′]≤ϵ₁, the probability of duplicating PUF P with a clone PUF P′, such that their output distributions are t-statistically close is less than some sufficiently small ϵ₁.

2. Unpredictable: Ideally, ^-PRED(κ₂):=Pr[r=r′], denoting the probability of the adversary guessing the correct response r of the PUF P to the challenge c, is negligible in κ₂ for all probabilistic polynomial time (PPT) adversaries . This refers to the decisional variant of the PUF unpredictability game described in Protocol 1.

3. Robust: Ideally, Pr[dist(y,z)>t|x←U_κ1, y∂P(x), z←P(x)]≤ϵ₂, i.e., the probability of a fixed PUF P yielding responses i-distant on the same input x is less than some sufficiently small ϵ₂. This property is satisfied by binding the PUF device d with a (m, , t, ϵ₃) fuzzy extractor (Gen, Rep).

4. Fuzzy Extraction: Ideally, during the enrollment phase for a noisy PUF P, given a challenge c, the PUF computes (R,H)←Gen(r), where r←P_d(c) and outputs H. The helper string H allows for R to be recovered when the challenge r′ is t-close to the original challenge r.

5. Indistinguishability: Ideally, the output of the PUF is computationally indistinguishable from a random string of the same length, such that the advantage of a PPT adversary is ^-IND()≤½+ϵ₃, where ϵ₃ is negligible. This requires that the adversary distinguishes between a random string, s, and the output of the fuzzy extractor, R_i, with probability at most negligibly more than ½. Thus, ^-IND−½≤ϵ₃, where ϵ₃ is negligible.

However, alternate definitions (e.g., Hori et al., “Quantitative and Statistical Performance Evaluation of Arbiter Physical Unclonable Functions on FPGAs,” 2010 International Conference on Reconfigurable Computing and FPCAs (ReConFig), pages 298-303, 2010: Maiti, A Systematic Approach to Design an Efficient Physical Undeniable Function, dissertation, Virginia Tech, 2012) for an ideal physical unclonable function may be substituted.Threshold Cryptography

Threshold cryptography involves distributing cryptographic operations among a set of participants such that operations are only possible with the collaboration of a quorum of participants. A trusted dealer generates a master asymmetric key pair ^pub, ^priv for the set of participants _i∈, ||=n. The private key is then split among the n participants, with each participant, receiving a share of ^priv. This constitutes a (t, n) sharing of ρ^priv, such that a quorum of at least t participants must combine their private shares in order to perform operations using the master private key.

While other secret schemes can be used with the present invention (e.g., Blakley, “Safeguarding cryptographic keys,” Proceedings of the 1979 AFIPS National Computer Conference, pages 313-317, AFIPS Press, 1979), an example will be described employing Shamir's polynomial interpolation construction (“How to Share a Secret,” Commun. ACM, volume 22, no. 11:612-613, 1979), which can be used for sharing a secret. A polynomial ƒ(⋅) of degree t−1 is defined, where the coefficients c_i remain private: ƒ(x)=c₀+c₁x+ . . . +c_t−1x^t−1 mod q. Without knowledge of the coefficients, ƒ(⋅) can be evaluated when at least t points of ƒ(⋅) are known by applying Lagrange's polynomial interpolation approach. A private key ^priv can be set as the free coefficient c₀(i.e., ƒ(0)=^priv), and a set of shares of the private key distributed to the participants (cf., e.g., Ertaul, “ECC Based Threshold Cryptography for Secure Data Forwarding and Secure Key Exchange in MANET (I),” NETWORKING 2005, Networking Technologies, Services, and Protocols; Performance of Computer and Communication Networks; Mobile and Wireless Communications Systems, volume 3462 of Lecture Notes in Computer Science, pages 102-113, Springer, 2005). To split the private key ^priv among n participants _i∈₁≤i≤n, the dealer computes _i's public, private key pair as r_i·G mod q,r_i such that r_i=ƒ(i), i≠0. Here, G∈E/_p is a base point of order q for elliptic curve E, and (P)_x (resp. (P)_y) refers to the x (resp. y) coordinate of point P on curve E. (The modulus that operations are performed under may be omitted where it is apparent from context). The public keys are made available to all participants, while the private keys are distributed securely to each participant (e.g., using the device's public key and Algorithm 3 below). All participants are also given access to (c_j·G)_{0≤j≤t−1}, which allows them to verify their secret key and the public keys of other participants by checking that:

$\begin{array}{cc}{r}_i·G=\sum _{j=0}^{t-1}{i}^j\left(c_j·G\right)& \mathrm{mod}p\end{array}$

This constitutes a (t,n) verifiable secret sharing (VSS) (e.g., Feldman, “A Practical Scheme for Non-interactive Verifiable Secret Sharing,” Proceedings of the 28th Annual Symposium on Foundations of Computer Science, SFCS '87, pages 427-438, IEEE Computer Society, 1987; Pedersen, “Non-Interactive and Information-Theoretic Secure Verifiable Secret Sharing,” Advances in Cryptology, CRYPTO 91, volume 576 of Lecture Notes in Computer Science, pages 129-140, Springer, 1992) of the private key ^priv, as participants are able to verify the legitimacy of their share with respect to a globally-known public key.

Now, given access to any t shares {(i,r_i)}1≤i≤t, where ƒ(⋅) has degree t−1 and t≤n, the shares (i, r_i) may be combined through Lagrange polynomial interpolation to evaluate ƒ(x):

$\begin{array}{cc}f\left(x\right)=\sum _{i=1}^t\left(r_i·\left(\prod _{\underset{j\ne i}{j=1}}^t\frac{x-x_j}{x_i-x_j}\right)\right)& \mathrm{mod}q\end{array}$ This allows any quorum of t participants _i∈⊆, |=t≤n to combine their shares {(i,r_i)}_1≤i≤t and recover the polynomial's free coefficient c₀=ƒ(0), which is the master asymmetric private key ^priv. Although we use the Lagrange form for the interpolating polynomial, other approaches (e.g., using a monomial basis or the Newton form) may be substituted. Similarly, although our exemplary construction prefers to evaluate ƒ(⋅) rather than recover the coefficients, the latter may be accomplished using a Vandermonde matrix representation and solving the system of linear equations.

FIG. 3-FIG. 6 illustrate the Lagrange polynomial interpolation of

$P\left(x\right)=-\frac{x^3}{6}+\frac{3x^2}{2}-\frac{7x}{3}+2.$ The interpolating polynomial P(⋅) is generated from a set of k points {(x_i, P(x_i))}_1≤i≤k. FIG. 3 illustrates the interpolating polynomial P(⋅)=1, which was generated from only a single point. FIG. 4 illustrates the interpolating polynomial P(⋅)=x, which was generated from two points. FIG. 5 illustrates the interpolating polynomial

$\stackrel{\_}{P}(·)=\frac{x^2}{2}-\frac{x}{2}+1,$ which was generated from three points. FIG. 6 illustrates the interpolating polynomial

$\stackrel{\_}{P}(·)=-\frac{x^3}{6}+\frac{3x^2}{2}-\frac{7x}{3}+2,$ which was generated from four points. As the degree of the polynomial is only three, any four points results in a perfect interpolation of the original polynomial. When the size of the set k exceeds the degree of the polynomial t−1 (i.e., k≥5), P(⋅) perfectly interpolates the original polynomial P(⋅). Thus in this example, the interpolating polynomial is generated from four points, which exceeds the degree (3) of the polynomial. Note that given any set of k<t points, no information about the secret P(0) is revealed, as an infinite number of polynomials of degree t−1 exist that satisfy the set of k<t points.

While an exemplary embodiment may use elliptic curve cryptography, it will be readily apparent that various other cryptographic frameworks (e.g., EIGamal, RSA, NTRU, etc.) could be employed. A number of threshold cryptographic operations can be carried out within this framework, using a variety of methods such as threshold encryption, decryption, and signing, threshold zero knowledge proofs of knowledge, threshold signcryption, and distributed key generation. Other elliptic curve mechanisms such as Massey-Omura, Diffie-Hellman, Menezes-Vanstone, Koyarna-Maurer-Okamoto-Vanstone, Ertaul, Demytko, etc. could likewise be employed.

An entity in possession of a device's enrollment information {_i^pub, c_i, helper_i} can thus encrypt a message m such that only the target device is able to recover it, using a method such as EIGamal encryption:

Algorithm 3 ElGamal Encryption
for Server s do
Lookup: challenge ci, generator G, modulus p and Helper hi
for Device pi
Select y ∈   p uniformly at random
Lookup public key A = pipriv • G mod p for Device pi
Generate shared key: y • G mod p
Encrypt message m by computing m + (yA)y mod q
Device pi ← {yG, m + (yA)y mod q, ci, G, p, hi}
end for

Then, if all participants of a group ⊆, where ||≥t, ||=n and t≤n, wish to decrypt an encryption (yG, m+(yrG)_x) of a message m∈[1, −1] using group private key r, threshold EIGamal decryption (e.g., per Ertaul) can be used as follows:

• Each participant _i∈ uses their secret key r_i=ƒ(i) to compute a shadow:

$\begin{array}{cc}{w}_i=\left(\prod _{\underset{j\ne i}{j=1}}^t\frac{-j}{i-j}\right)·r_i& \mathrm{mod}q\end{array}$

• Each participant then broadcasts their partial decryption S_i defined as S_i=w_i·yG mod q.
• Each participant locally computes the value:

$\begin{array}{c}S=\begin{array}{cc}\sum _{i=1}^t{S}_i& \mathrm{mod}q\end{array}\\ =\left(\sum _{i=1}^t{w}_i\right)·\mathrm{yG}\\ =r·\mathrm{yG}\end{array}$

• Finally, each participant may now locally recover the message m by computing(m+(yrG)_y)−S mod q=(m+(yrG)_y)−(ryG)_y=m

Likewise, a group ⊆ where ||≥t, ||=n and t≤n, can use a threshold signature scheme (e.g., Chen et al., “An efficient threshold group signature scheme,” IEEE Region 10 Conference TENCON, volume B. pages 13-16 Vol. 2, 2004; Hua-qun et al., “Verifiable (t, n)Threshold Signature Scheme based on Elliptic Curve,” Wuhan University Journal of Natural Sciences, volume 10, no. 1:165-168, 2005; Ibrahim et al., “A Robust Threshold Elliptic Curve Digital Signature providing a New Verifiable Secret Sharing Scheme,” IEEE 46th Midwest Symposium on Circuits and Systems, volume 1, pages 276-280 Vol. 1, 2003; Kim et al., “Threshold Signature Schemes for ElGamal Variants,” Computer Standards and Interfaces, volume 33, no. 4:432-437, 2011; Shao, “Repairing Efficient Threshold Group Signature Scheme,” International Journal of Network Security, 2008) to generate a signature representing all of for message m as follows:

• Each participant _i∈ uses their secret key r_i=ƒ(i) and a random integer y_i∈_q to calculate their individual signature (R_i, S_i) for message m.
  • First, R_i is computed and publicized to all participants _i∈:R_i=y_i·G mod p
  • Next, each participant _i calculates R, e, S_i as follows:

$\begin{array}{cc}R=\sum _{i=1}^t{R}_i& \mathrm{mod}p\end{array}$ $e=h\left(\begin{array}{cc}m,{\left(R\right)}_y& \mathrm{mod}q\end{array}\right)$ $\begin{array}{cc}{S}_i=y_i+r_{i}e\left(\prod _{\underset{j\ne i}{j=1}}^t\frac{-j}{i-j}\right)& \mathrm{mod}q\end{array}$

• •  h(⋅) or H(⋅) denotes a cryptographic hash function. Each participant broadcasts S_i to an appointed secretary (for convenience, and who need not be trusted).
• The secretary, having received all (R_i, S_i) pairs, verifies the signature by computing:

$\begin{array}{cc}R=\sum _{i=1}^t{R}_i& \mathrm{mod}p\end{array}$ $e=h\left(m,{\left(R\right)}_y\mathrm{mod}q\right)$ $\begin{array}{cc}{R}_i=S_i·G+e\left(\prod _{\underset{j\ne i}{j=1}}^t\frac{-j}{i-j}\right)\left(-p_i^{\mathrm{pub}}\right)& \mathrm{mod}p\end{array}$ If constructed properly, this equation will hold as:

$\begin{array}{c}{R}_i\stackrel{?}{=}{S}_i·G+e\left(\prod _{\underset{j\ne i}{j=1}}^t\frac{-j}{i-j}\right)\left(-p_i^{\mathrm{pub}}\right)\mathrm{mod}p\\ =\left(y_i+r_{i}e\left(\prod _{\underset{j\ne i}{j=1}}^t\frac{-j}{i-j}\right)\right)·G+e\left(\prod _{\underset{j\ne i}{j=1}}^t\frac{-j}{i-j}\right)\left(-r_{i}G\right)\\ =y_{i}G+e{r}_{i}G\left(\prod _{\underset{j\ne i}{j=1}}^t\frac{-j}{i-j}\right)+e\left(-r_{i}G\right)\left(\prod _{\underset{j\ne i}{j=1}}^t\frac{-j}{i-j}\right)\\ =y_{i}G+\left(e{r}_{i}G-e{r}_{i}G\right)\left(\prod _{\underset{j\ne i}{j=1}}^t\frac{-j}{i-j}\right)\\ =y_{i}G\\ :=R_i\end{array}$

• If these hold, the secretary calculates:

$\begin{array}{cc}S=\sum _{i=1}^t{S}_i& \mathrm{mod}q\end{array}$ which computes the group signature ((R)_y mod q, S) over m.

• Upon receipt of (R,S), receiver _R checks its validity against the public key ^pub of the entire group of participants _i∈_1≤i≤n:eh(m,(S·G+ϵ·−^pub)_y mod q)which holds on valid signatures because:

$\begin{array}{c}e\stackrel{?}{=}\begin{array}{cc}h(m,{\left(S·G+e·-{𝒫}^{\mathrm{pub}}\right)}_y& \mathrm{mod}q)\end{array}\\ =h\left(m,{\left(\left(y+\mathrm{re}\right)·G+e·\left(-\mathrm{rG}\right)\right)}_y\right)\\ =h\left(m,{\left(\mathrm{yG}+\mathrm{erG}-\mathrm{erG}\right)}_y\right)\\ =h\left(m,{\left(\mathrm{yG}\right)}_y\right)\\ =\begin{array}{cc}h(m,{\left(R\right)}_y& \mathrm{mod}q)\end{array}\end{array}$

The participants of a group ⊆ where ||≥t, ||=n and t≤n can also collaborate to demonstrate possession of a shared private key ^priv=r∈[1, q−1] using a threshold Zero Knowledge Proof of Knowledge (e.g., Sardar et al., “Zero Knowledge Proof in Secret Sharing Scheme Using Elliptic Curve Cryptography,” Global Trends in Computing and Communication Systems, volume 269 of Communications in Computer and Information Science, pages 220-226, Springer, 2012) as follows:

• The group public key is ^pub=rG, where r is a shared secret and G is a group generator. The verifier chooses an ephemeral nonce N and distributes this to all participants of .
• Each participant _i∈ uses their secret share r_i=ƒ(i) and a random nonce integer y_i to calculate their individual proof (B_i, M_i) of the shared secret r.
  • First, B_i is computed and publicized to all participants _i∈:B_i=y_i·G mod
  • Each participant locally computes:

$\begin{array}{cc}B=y·G=\sum _{i=1}^t{B}_i& \mathrm{mod}p\end{array}$

• • Next, each participant _i calculates e, M_i as follows:

$e=h\left(G,B,{𝒫}^{\mathrm{pub}},N\right)$ $\begin{array}{cc}{M}_i=y_i+r_{i}e\left(\prod _{\underset{j\ne i}{j=1}}^t\frac{-j}{i-j}\right)& \mathrm{mod}q\end{array}$

• Upon receipt of (B_i, M_i)_1≤i≤t, the verifier computes:

$\begin{array}{cc}B=\sum _{i=1}^t{B}_i& \mathrm{mod}p\end{array}$ $\begin{array}{cc}M=\sum _{i=1}^t{M}_i& \mathrm{mod}q\end{array}$ $e=h\left(G,B,{𝒫}^{\mathrm{pub}},N\right)$

• Next, the verifier checks the proof's validity against the public key ^pub=rG.

$\begin{array}{c}B\stackrel{?}{=}\begin{array}{cc}M·G-e·{𝒫}^{\mathrm{pub}}& \mathrm{mod}p\end{array}\\ =\left(y+\mathrm{re}\right)·G-e·\left(\mathrm{rG}\right)\\ =\mathrm{yG}+\mathrm{reG}-\mathrm{reG}\\ =\mathrm{yG}\end{array}$

If B=M·G−e·^pub, the verifier accepts the threshold zero knowledge proof as valid, and rejects the proof otherwise.

The process of signcrypting (e.g., Changgen et al., “Threshold Signcryption Scheme based on Elliptic Curve Cryptosystem and Verifiable Secret Sharing,” International Conference on Wireless Communications, Networking and Mobile Computing, volume 2, pages 1182-1185, 2005: Zheng, “Digital Signcryption or How to Achieve Cost (Signature & Encryption)«Cost (Signature)+Cost(Encryption),” Advances in Cryptology, CRYPTO '97, volume 1294 of Lecture Notes in Computer Science, pages 165-179, Springer, 1997; Zheng et al., “How to Construct Efficient Signcryption Schemes on Elliptic Curves,” Inf. Process. Lett., volume 68, no. 5:227-233, 1998) a message facilitates performing both signing and encrypting a message at a cost less than computing each separately. Given a message m∈[1,q −1] and a receiver _R with public key _R^pub, signcryption can be generated as follows:

• Each _i∈ selects a random k_i∈[1,q−1] and computes Y_i=k_i·G and publicly broadcasts this to both a secretary (for convenience, and who need not be trusted) and the receiver _R. Each _i∈ also computes Z_i=k_i·_R^pub which is privately (for example, using Algorithm 3) sent to _R.
• The secretary computes:

$Z=\sum _{i=1}^t{Z}_i=\sum _{i=1}^t{k}_i·p_R^{\mathrm{pub}}=k·p_R^{\mathrm{pub}}$ $\begin{array}{cc}r=m·{\left(Z\right)}_x& \mathrm{mod}p\end{array}$ and broadcasts r (not to be confused with r_i, participant _i's share of ^priv) to each signer _1≤i≤t.

• Each signer _1≤i≤t computes:

$\begin{array}{cc}{x}_i=\prod _{\underset{j\ne i}{j=1}}^t\frac{-j}{i-j}& \mathrm{mod}q\end{array}$ $\begin{array}{cc}{e}_i=r_i·x_i& \mathrm{mod}q\end{array}$ $\begin{array}{cc}{s}_i=k_i-e_i·r& \mathrm{mod}q\end{array}$ where r_i=ƒ(i) is _i's share of ^priv. Each signer sends their partial signcryption s_i to the secretary.

• Upon receipt of a partial signcryption s_i, the secretary computes Y′_i=r·x_i·_i^pub+s_i·G in order to verify the validity of the partial signcryption by checking Y_iY′_i.
• Once receiving all partial signcryptions s_i and checking their validity, the secretary combines them to compute:

$\begin{array}{cc}s=\sum _{i=1}^t{s}_i& \mathrm{mod}q\end{array}$

•  and (r, s) is the final signcryption sent to receiver _R.
• The receiving participant _R, which has now received {Y_i=k_i·G}_{i∈[1 . . . n]}, (r, s), computes:

$Y=\sum _{i=1}^t{Y}_i=\sum _{i=1}^t\left(k_i·G\right)=k·G$ $Y^{\prime }=r·{𝒫}^{\mathrm{pub}}+s·G$ $Z^{\prime }=p_R^{\mathrm{priv}}·Y^{\prime }$

• The recipient _R then verifies that:

$\begin{array}{c}{Y}^{\prime }\stackrel{?}{=}r·{𝒫}^{\mathrm{pub}}+s·G\\ =r·{𝒫}^{\mathrm{pub}}+\sum _{i=1}^t{s}_i·G\\ =r·{𝒫}^{\mathrm{pub}}+G·\sum _{i=1}^t\left(k_i-e_i·r\right)\\ =r·{𝒫}^{\mathrm{pub}}+\left(G·\sum _{i=1}^t{k}_i\right)-\left(G·\sum _{i=1}^t{e}_i·r\right)\\ =r·{𝒫}^{\mathrm{pub}}+k·G-r·G·\sum _{i=1}^t{e}_i\\ =r·{𝒫}^{\mathrm{pub}}+k·G-r·G·\left(\sum _{i=1}^t{r}_i·\left(\prod _{\underset{j\ne i}{j=1}}^t\frac{-j}{i-j}\right)\right)\\ =r·{𝒫}^{\mathrm{pub}}+k·G-r·G·\left(f\left(0\right)\right)\\ =r·{𝒫}^{\mathrm{priv}}·G+k·G-r·G·{𝒫}^{\mathrm{priv}}\\ =k·G=Y\end{array}$

•  If these hold, the group signature over m is valid.
• The recipient _R can now recover the message m by computing:

$\begin{array}{c}r·{\left(Z^{\prime }\right)}_x^{-1}=\left(m·{\left(Z\right)}_x\right)·{\left(Z^{\prime }\right)}_x^{-1}\\ =\left(m·{\left(k·p_R^{\mathrm{pub}}\right)}_x\right)·{\left(p_R^{\mathrm{priv}}·Y^{\prime }\right)}_x^{-1}\\ =\left(m·{\left(k·\left(p_R^{\mathrm{priv}}·G\right)\right)}_x\right)·{\left(p_R^{\mathrm{priv}}·\left(k·G\right)\right)}_x^{-1}\\ =m\end{array}$ With this, the recipient _R has both verified the group's signature over message m, as well as decrypted m.Distributed Key Generation

Standard threshold cryptographic operations (e.g., those discussed above) traditionally require the presence of a trusted dealer to define a generating polynomial ƒ(⋅), select a secret r, and distribute shares of r to all participants _i∈. Distributed key generation protocols (e.g., Ibrahim; Pedersen, “A Threshold Cryptosystem without a Trusted Party,” Advances in Cryptology, EURO CRYPT 91, volume 547 of Lecture Notes in Computer Science, pages 522-526, Springer, 1991; Tang, “EGDKG: A Distributed Key Generation Protocol Based on Elliptic Curve Discrete Logarithm,” Technical Report 04-838, Department of Computer Science, University of Southern California, 2004) remove the necessity of a trusted dealer, and allow a set of participants to generate shares of a secret where no one knows the shared secret r. This can be accomplished in the present context as follows:

• Each participant _i∈ defines a random polynomial ƒ_i(⋅) of degree t−1, where t is the threshold. The temporary private value of participant _i is c₀⁽ⁱ⁾, the free coefficient of ƒ_i(⋅).
• Each participant _i∈ privately sends ƒ_i(j) to participant _j, ∀j∈[1, . . . , n]/i.
• Participant _i broadcasts {c_k⁽ⁱ⁾·G mod p}_{k∈[0, . . . ,t−1]}, commitments to the coefficients of ƒ_i(⋅).
• Participant _i broadcasts {ƒ_i(j)·G mod p}_{j∈[0, . . . , n]}, the public shares for all participants.
• Each participant _j≠i∈ must now verify the shares they have received.
  • First, each participant _j≠i verifies that:

$\begin{array}{c}{f}_i\left(j\right)·G\stackrel{?}{=}\begin{array}{cc}\sum _{k=0}^{t-1}{j}^k\left(c_k^{\left(i\right)}·G\right)& \mathrm{mod}p\end{array}\\ =G·\left(\sum _{k=0}^{t-1}{j}^k{c}_k^{\left(i\right)}\right)\\ =G·f_i\left(j\right)\end{array}$

• • Similarly, each participant _j≠i∈ verifies that their share is consistent with other shares:

$\begin{array}{c}{c}_0^{\left(i\right)}·G\stackrel{?}{=}\begin{array}{cc}\sum _{i=1}^t\left(\left(\prod _{\underset{j\ne i}{j=1}}^t\frac{-j}{i-j}\right)f_i\left(j\right)·G\right)& \mathrm{mod}p\end{array}\\ =G·\sum _{i=1}^t\left(\left(\prod _{\underset{j\ne i}{j=1}}^t\frac{-j}{i-j}\right)f_i\left(j\right)\right)\\ =G·c_0^{\left(i\right)}\end{array}$

• If these two verifications succeed, each participant _i∈ computes its share of the master asymmetric private key r:

$\begin{array}{cc}{r}_i=\sum _{j=1}^n{f}_j\left(i\right)& \mathrm{mod}q\end{array}$

• Similarly, the master asymmetric public key for the group is computed as:

$\begin{array}{cc}r·G=\sum _{j=1}^n\sum _{i=1}^n\left(f_j\left(i\right)·G\right)& \mathrm{mod}p\end{array}$

The distributed key generation protocol is preferably secure against an adversary that attempts to bias the output distribution, as in the attack described by Gennaro et al. (“Secure Distributed Key Generation for Discrete-Log Based Cryptosystems,” Advances in Cryptology, EUROCRYPT 99, volume 1592 of Lecture Notes in Computer Science, pages 295-310, Springer, 1999). (Gennaro et al. (“Secure Applications of Pedersen's Distributed Key Generation Protocol,” Topics in Cryptology, CT-RSA 2003, volume 2612 of Lecture Notes in Computer Science, pages 373-390, Springer, 2003) later concluded that many threshold operations may be performed securely despite an adversary's ability to bias the output distribution). Similarly, threshold constructions are preferably secure against both static as well as adaptive malicious adversaries (Abe et al., “Adaptively Secure Feldman VSS and Applications to Universally-Composable Threshold Cryptography,” Advances in Cryptology, CRYPTO 2004, volume 3152 of Lecture Notes in Computer Science, pages 317-334, Springer, 2004; Jarecki et al., “Adaptively Secure Threshold Cryptography: Introducing Concurrency, Removing Erasures,” Advances in Cryptology, EUROCRYPT 2000, volume 1807 of Lecture Notes in Computer Science, pages 221-242, Springer, 2000; Libert et al., “Adaptively Secure Forward-Secure Non-interactive Threshold Cryptosystems,” Information Security and Cryptology, volume 7537 of Lecture Notes in Computer Science, pages 1-21, Springer, 2012).

PUF-Enabled Threshold Cryptography

The core functionality of a PUF is extracting a unique mapping between the challenge (input) domain and the response (output) range. As the mapping from challenges to responses is unique for each PUF-enabled device, collecting a set of challenge-response pairs (CRPs) through a provisioning process allows the device to be verified in the future. Protocol 3 illustrates the naïve provisioning process that underlies many PUF-enabled protocols.

Protocol 3: Challenge-Response Provisioning
	PUF Device D		Server s
		←	challenge c ∈ {0, 1}κ1
	P(c)     r ∈ {0, 1}κ2	→	store(D, {c, r})

Authentication proceeds by issuing a challenge for which the response is known to the server, and verifying that the response is i-close to the expected response. However, this lightweight naïve protocol has many limitations. During enrollment, a large number of challenge-response pairs must be collected, as each pair can only be used once for authentication. If an adversary observed the response, it could masquerade as the device. Similarly, the challenge-response database is sensitive, as an adversary could apply machine learning to fully characterize the PUF mapping [Rührmair I]. These issues can be entirely eliminated by applying cryptographic constructs around the PUF functionality.

In the example of an embodiment employing elliptic curve cryptography, Algorithms 4 and 5 below can be used to allow a PUF-enabled device to locally store and retrieve a sensitive value without storing any sensitive information in non-volatile memory. Algorithm 4 illustrates the storing of a sensitive value _i using a PUF, and Algorithm 5 illustrates the dynamic regeneration of _i. The challenge c_i and helper data helper_i can be public, as neither reveals anything about the sensitive value _i. While the present example uses encryption of _i by exclusive-or, ⊕, _i could also be used as a key to other encryption algorithms (e.g., AES) to enable storage and retrieval of arbitrarily sized values.

Algorithm 4 PUF-Store
Goal: Store value  i
for PUF Device d do
Select finite field  n of order n
Select E, an elliptic curve over  n
Find G ∈ E/ n, a group generator
Select challenge ci ∈  n
x = H(ci, E, G, n)
O = PUF(x)
helperi = Pi = O ⊕ ECC( i)
Write {ci, helperi} to non-volatile memory
end for

Algorithm 5 PUF-Retrieve
Goal: Retrieve value  i
for PUF Device d do
Read {ci, helperi} from non-volatile memory
x ← H(ci, E, G, n)
O′ = PUF(x)
i ← D((ECC( i) ⊕ O) ⊕ O′)
end for

Whenever O and O′ are t-close, the error correcting code ECC can be passed to a decoding algorithm D which will recover the sensitive value _i.

Using Algorithm 6, a local device can perform an enrollment protocol using the PUF.

This allows each PUF circuit to generate a local public key _i^pub, which is useful for bootstrapping more complex key setup algorithms (e.g., the distributed key generation protocol in Algorithm 7). When the key setup algorithm is performed internal

Algorithm 6 The Enrollment Algorithm
for Device d do
ci ∈  p, a group element
x = H(ci, E, G, p, q)
O = PUF(x)
helperi = O ⊕ ECC  ipriv mod q)
i              pub = Ai =  ipriv · G mod p
Store { ipub, ci, helperi}
end for

to the device (rather than externally among a set of distinct devices), this bootstrap process may not be necessary.

In accordance with the invention, PUF-based cryptographic primitives are adapted to secret sharing to permit threshold cryptography founded on PUF or other root of trust. Using the example of an embodiment employing elliptic curve cryptography, distributed key generation is used to generate a number of shares (for example, two: r₁, r₂) of a master private key ^priv=(r₁+r₂) mod q), which itself is never generated or constructed. (It is also possible to work directly with a message (e.g., as described by Ertaul) rather than a private key). The protocol is summarized in Algorithm 7: PUF-DKG, where an exemplary implementation would choose (t,n) as (2,2).

Algorithm 7 PUF-DKG
Goal: Generate shares of master private key   priv
for 1 ≤ i ≤ n do
Select random polynomial fi(·) = c0(i) + . . . + ct−1(i)xt−1 mod q
Compute fi(j), ∀j ∈ [1, . . . , n]/i
Store coefficient commitments {ck(i) · G mod p}k∈[0,...,t−1]
Store share commitments {fi(j) · G mod p}j∈[0,...,n]
for 1 ≤ i ≤ n do
Verify
fi⁡(j)·G⁢=?⁢∑k=0t-1⁢⁢jk⁡(ck(i)·G)⁢⁢mod⁢⁢p
Verify
c0(i)·G⁢=?⁢∑i=1t⁢⁢((∏j=1j≠it⁢⁢-ji-j)⁢⁢fi⁡(j)·G)⁢⁢mod⁢⁢p
end for
Recover share
ri=∑j=1n⁢⁢((∏j=1j≠it⁢⁢-ji-j)⁢⁢fj⁡(i))⁢⁢mod⁢⁢q
Recover public key
𝒫pub=r·G=∑j=1n⁢⁢∑i=1n⁢⁢(((∏j=1j≠it⁢⁢-ji-j)⁢⁢fj⁡(i))·G)⁢⁢mod⁢⁢p
end for

Using Algorithms 4 and 5 for storing and retrieving a sensitive value, and Algorithm 7 for performing the initial distributed key generation protocol, arbitrary PUF-enabled threshold cryptographic operations (e.g., decryption, digital signatures, zero knowledge proofs) can now be performed. Algorithm 8 describes how to evaluate an arbitrary threshold cryptographic operation that requires as input a participant's share r_i. Note that the recovered share r_i has already been multiplied by the Lagrange terms

$\left(\prod _{j=1,j\ne i}^t\frac{-j}{i-j}\right).$

Algorithm 8 PUF-Threshold-OP
Goal: Perform threshold operation
Assume: PUF-DKG (Algorithm 7) has been executed by PUF Device d
for Server s do
Issue Command     and Auxiliary information Aux
end for
for PUF Device d do
for each challenge ci (Challenge c = c0∥ . . . ∥cn) do
Read challenge ci and helper data hi
Recover share ri ←PUF-Retrieve(ci, hi)
Perform threshold. operation    (ri, Aux)
end for
Combine threshold operations     ← Combine({   (ri, Aux)}0≤i≤n)
return Result
end for
for Server s do
Process operation
end for

This enables any threshold cryptographic operation (e.g., decryption, digital signature generation, zero knowledge proofs) to be performed by a PUF-enabled participant without ever generating, reconstructing, or storing their private key. Further, from an external perspective (e.g., the server), the PUF-enabled device simply implements standard public key cryptographic protocols. That is, the server never issues a challenge or stores helper data, and its interaction with the device is indistinguishable from any standard public key cryptography device.

By internalizing the challenge-response functionality of the PUF, and utilizing Algorithms 4 and 5 to locally store and recover a value (e.g., a cryptographic key), arbitrary (e.g., symmetric or asymmetric) cryptographic operations can be performed without need for issuing or storing auxiliary (e.g., challenges or helper data) information. While one embodiment described herein advantageously strengthens the construction through both distributed key generation and threshold cryptography, neither is necessary to support arbitrary cryptographic operations through localized storage and retrieval of a value using a device's PUF functionality according to the present invention.

Although threshold cryptography typically considers distributing operations across physically-distinct nodes, in one embodiment of the present invention, threshold cryptography may be applied within a single device. As an example, a device may be equipped, e.g., with two PUF circuits and provided with the ability to execute at least two instructions at the same time (e.g., through multiple CPU cores). An embodiment of the present invention with only one PUF circuit would simply execute operations over each share sequentially, rather than querying the multiple PUF circuits in parallel. FIG. 2 illustrates a device equipped with two PUF circuits to enable local threshold cryptographic operations; the device may be, for example, an FPGA with a separate core containing each PUF. The potentially extractable output of a single PUF may then be obviated by constructing a local (2,2) threshold system with each of the parts _i acting as a distinct participant. For example, each part may select a random challenge, run the enrollment algorithm (Algorithm 6) to generate an asymmetric key pair _i^pub=_i^priv·G, _i^priv and locally store its public enrollment information and then together run the distributed key generation protocol (Algorithm 7) and perform all cryptographic operations over a private key that is never actually constructed. When threshold cryptography is applied within a single device, it may not be necessary to run the enrollment algorithm (Algorithm 6) to generate an asymmetric key pair as all computations are performed internal to the device.

Algorithm 9 describes how a dual-PUF device can compute cryptographic operations in a threshold manner by constructing a (2, 2) threshold sharing within the device using distributed key generation. That is, the two parts establish a private key known to neither part through distributed key generation and publicize the corresponding public key ^pub. All operations targeted at. the device are now performed in a threshold manner through internal collaboration (with each part retrieving its share r_i and performing a local threshold operation, and the results are combined to complete a threshold operation ), while the input/output behavior of the device remains unchanged to external systems.

Algorithm 9 Dual-PUF-Threshold-OP
Goal: Perform threshold operation     at time τ
One-Time Setup Stage
for each PUF Core pi ∈     do
Run Algorithm 6: Enrollment, Publicize  ipub
end for
Run(2, 2)-PUF-DKG Protocol, Publicize    pub
Evaluation Stage
for each PUF Core pi ∈     do
Recover share ri(τ) ←PUF-Retrieve (ci(τ), helperi(τ))
i              (                            )  ←    (ri(τ)), PUF core local threshold share
end for
return     ← Combine({   ,   })

Thus, rather than being constrained to a mapping between a challenge issued to the device and its response (which to an extent may be a function of the challenge), a multi-PUF device d_i can have a single static external identity, _i^pub. The challenge-response functionality of each PUF core is used to maintain each share of the device's private identity, _i^priv, which is never generated or constructed. This renders a side channel attack more difficult for a remote adversary, which now must observe and resolve multiple values simultaneously generated within the device. Each part retrieves its share r_i^(τ) and performs a local threshold operation, and the shares are combined to complete the operation .

Referring to FIG. 7 and FIG. 8, the core operations of an example embodiment employing elliptic curve cryptography, division of a key into two shares, and a (2, 2) threshold operation, are described.

• Enrollment Command 1: During the initial enrollment process, the server and device agree on an elliptic curve E defined over a finite field and base point G of order q, where is λ bits long. The server issues the enrollment command to the device.
• Distributed Key Generation 2: The device performs a distributed key generation locally, creating shares (r₀,r₁) of the master private key (which is never generated or constructed) and its public key A=(r₀+r₁)·G. Rather than directly add the shares together (which would construct the private key r=r₀+r₁), the public key is formed by computing (r₀·G)+(r₁·G).
• Helper Data Generation 3: The device generates a random challenge c=c₀∥c₁, where ∥ denotes concatenation and each c_i block is λ bits long. The device links each share r_i to the output O_i of the PUF on challenge c_i through fuzzy extraction, which outputs a public helper h_i. As the PUF output O_i is noisy, there is no guarantee that when queried on challenge c_i in the future, the new output O′_i will satisfy O′_i=O_i. However, it is assumed that O_i and O′_i will be t-close with respect to some distance metric (e.g. Hamming distance). Thus, an error correcting code may be applied to the PUF output such that at most t errors will still recover O_i. Error correction may be applied over each share r_i, and this value blinded with the output of the PUF O_i on challenge c_i, so that each helper value h_i=ECC(r_i)⊕O_i reveals no information about share r_i. During recovery through fuzzy extraction, computing the exclusive-or of ECC(r_i)⊕O_i⊕O′_i will return r_i whenever O_i and O′_i are t-close. The device locally stores the challenge c=c₀∥c₁ and helper data h=h₀∥h₁, which will allow it to later recover the shares. Note that both the challenge and the helper data are public, and reveal nothing about the shares or the device's private key without invoking the PUF. This process is described by Algorithm 4.
• Returned Public Key 4: The device returns its public enrollment information {A=(r₀+r₁)·G} to the server.
• Store Enrollment 5: The server stores the device's public enrollment information along with a (non-sensitive) identifier that is unique to the device (e.g., a serial number).
• Threshold Operation Query 6: When the server wishes the device to perform a cryptographic operation (e.g., decryption, digital signature generation, zero knowledge proof authentication), it issues:
  • the appropriate command for the operation to be performed
  • any auxiliary data Aux that is necessary for the operation (e.g., ciphertext to be decrypted, a message to be signed)
• PUF Retrieval 7: The device reads the challenge c=c₀∥c₁ and helper data h=h₀∥h₁ from its local storage. The device then queries the PUF on each challenge block c_i, and combines the output O′_i with the helper block h_i and error correcting code to recover each share block r_i. This process is described by Algorithm 5.
• Threshold Operation 8: The device performs a threshold operation (r_i, Aux) over each share r_i. Algorithm 8 describes this process for any arbitrary threshold operation .
• Combined Threshold Operations 9: The device combines the threshold operations to form the complete operation and returns the result to the server.
• Process Operation 10: The server finally performs any additional processing required for the operation (e.g., verifying a zero knowledge proof).Share Refreshing

Various share refresh protocols (e.g., Frankel et al., “Optimal-Resilience Proactive Public-Key Cryptosystems,” 38th Annual Symposium on Foundations of Computer Science, pages 384-393, 1997; Herzberg et al., “Proactive Public Key and Signature Systems,” Proceedings of the 4th ACM Conference on Computer and Communications Security, CCS '97, pages 100-110, ACM, 1997; Herzberg et al., “Proactive Secret Sharing Or: How to Cope With Perpetual Leakage,” Advances in Cryptology, CRYPTO 95, volume 963 of Lecture Notes in Computer Science, pages 339-352, Springer, 1995) allow each of a set of players _i∈ to refresh their share r_i^(τ) of an original secret r at time period τ into a new share r_i^(τ+1) such that the resulting set of new shares {r_i^(τ+1)}_{i∈[1 . . . n]} remains a sharing of the original secret. This protocol does not require reconstruction of the master secret r, so a mobile adversary would have to compromise t players in a fixed time period τ in order to recover the shared secret. Assuming a polynomial ƒ(⋅) of degree (t−1) represents a shared secret r=ƒ(0) amongst n participants each having a share r_i=ƒ(i), and denoting encrypting for player _j as ENC_j(⋅) and decryption by _j as DEC_j(⋅), the set of players _i∈ can refresh their sharing of r using such a protocol as follows:

• Each player _i defines a new polynomial of degree (t−1) such that δ_i(0)=0:δ_i(⋅)=Δ₁⁽ⁱ⁾x+ . . . +Δ_m⁽ⁱ⁾x^t−1
•  where the set {Δ_m^(i)}_{m∈[1 . . . t−1]} chosen randomly from _q.
• Each player _i computes the following sets:{ϵ_im=Δ_m⁽ⁱ⁾·G}_{m∈[1 . . . t−1]}{u_ij=δ_i(j)}_{j∈[1 . . . n]}{e_ij=ENC_j(u_ij)}_{j∈[1 . . . n]}
•  and broadcasts a verifiable secret sharing VSS_i^(τ)=i, τ, {ϵ_im}, {e_ij} and their signature SIG_i(VSS_i^(τ)).
• Each player _i recovers u_ji=DEC_i(ENC_i(u_ji)) and verifies ∀j≠i:

$\begin{array}{c}{u}_{\mathrm{ji}}·G\stackrel{?}{=}\sum _{k=1}^{t-1}{i}^k{ϵ}_{\mathrm{jk}}\\ =\sum _{k=1}^{t-1}\left(i^k{\Delta }_k^{\left(j\right)}·G\right)\\ =G·\sum _{k=0}^{t-1}{i}^k{\Delta }_k^{\left(j\right)}\\ =G·{\delta }_j\left(i\right)=G·u_{\mathrm{ji}}\end{array}$

• Finally, each player _i updates their share from time period (τ) as:

$r_i^{\left(\tau +1\right)}\leftarrow r_i^{\left(\tau \right)}+\left(\begin{array}{cc}\sum _{j=1}^n{u}_{\mathrm{ji}}& \mathrm{mod}q\end{array}\right)$ Thus, the refreshed set of shares {r_i^(τ+1)}_{i∈[1 . . . n]} remains a sharing of the master private key ^priv, and yet knowledge of t−1 or fewer shares from time period τ is useless in time period τ+1.

As outlined in Algorithm 10, participants can update their share r_i^(τ) in time period τ to a new share r_i^(τ+1) in the next time period such that the set of shares {r_i}_{i∈[1 . . . n]} remains a sharing of the master private key ^priv.

Algorithm 10 PUF-Share-Update
Goal: Generate new share ri(τ+1)
for all Participants pi ∈   do
Recover share ri(τ)
ri(τ) ← PUF-Retrieve(ci(τ), helperi(τ))
Select random polynomial of degree (t − 1) such that δi(0) = 0:
δi(·) = Δ1(i)x + . . . + Δm(i)xt−1
Compute
{ϵim = Δm(i) · G}m∈[1...t−1]
{uij = δi(j)}j∈[1...n]
{eij = ENCj(uij)}j∈[1...n]
Broadcast
VSSi(τ) =   i, τ, {ϵim}, {eij}   , SIGi(VSSi(τ))
Verify ∀j ≠ i
DECi⁡(ENCi⁡(uji))·G⁢=?⁢∑k=1t⁢⁢ik⁢ϵjk
Update share as:
ri(τ+1)←ri(τ)+(∑j=1n⁢⁢uji⁢⁢mod⁢⁢q)
Store ri(τ+1) and update PUF challenge:
{ci(τ+1), helperi(τ+1)} ← PUF-Store(ri(τ+1))
end for

The hardware device performs Algorithm 10 at Share Refresh 11 in FIG. 8 to generate new shares r_i^(τ+1) for the next time period τ+1. At PUF Refresh and Store 12, the hardware device generates a new challenge c_i^(τ+1), which will refresh the challenge-helper pair for the next time period. The hardware device uses the new challenge to store the updated share r_i^(τ+1). Algorithms 8 and 9 are modified to refresh both the threshold shares as well as the challenge-helper pair, with Algorithms 11 and 12, respectively, reflecting the modifications.

Algorithm 11 PUF-Threshold-OP-Refreshing
Goal: Perform threshold operation
Assume: PUF-DKG (Algorithm 7) has been executed by PUF Device d
for Server s do
Issue Command     and Auxiliary Information Aux
end for
for PUF Device d do
for each challenge ci (Challenge c = c0∥ . . . ∥cn) do
Read challenge ci and helper data hi
Recover share ri ←PUF-Retrieve(ci, hi)
Perform threshold operation    (ri, Aux)
end for
Combine threshold operations     ← Combine({   (ri, Aux)}0≤i≤n)
return Result
for each share ri do
Update share
ri(τ+1) ← PUF-Share-Update(ri(τ))
Store ri(τ+1) and update PUF challenge:
{ci(τ+1), helperi(τ+1)} ← PUF-Store(ri(τ+1))
end for
end for
for Server s do
Process operation
end for

Algorithm 12 Dual-PUF-Threshold-OP-Refreshing
Goal: Perform threshold operation     at time τ
One-Time Setup Stage
for each PUF Core pi ∈     do
Run Algorithm 6: Enrollment, Publicize    ipub
end for
Run (2,2)-PUF-DKG Protocol, Publicize    pub
Evaluation Stage
for each PUF Core    i  ∈     do
Recover share ri(τ) ←PUF-Retrieve(ci(τ) , helperi(τ))
←    (ri(τ)), PUF core local threshold share
Update share
ri(τ+1) ← PUF-Share-Update(ri(τ))
Store ri(τ+1) and update PUF challenge:
{ci(τ+1), helperi(τ+1)} ← PUF-Store(ri(τ+1))
end for
return     ← Combine({   ,   })

Referring for example to a single-PUF embodiment as shown in FIG. 1, Algorithm 10 can be logically split into a preparation and application phase. During preparation, each participant generates their random polynomial and distributes their portion of the update to the other participants. After each participant has broadcast their portion of the share update, the preparation phase is complete. Algorithm 13 illustrates the preparation phase of share updating. Note that in one embodiment, the broadcast phase may be omitted when Algorithm 13 is applied within a single device (e.g., an FPGA).

Algorithm 13 PUF-Share-Update-Preparation
Goal: Prepare update to construct share ri(τ+1)
for all Participants    i ∈     do
Select random polynomial of degree (t − 1) such that δi(0) = 0:
δi(·) = Δ1(i)x + . . . + Δm(i)xt−1
Compute
{∈ im = Δm(i) · G}m∈[1...t−1]
{μij = δi(j)}j∈[1...n]
{eij = ENCj(uij)}j∈[1...n]
Broadcast
VSSi(τ) =    i, τ, {∈ im}, {eij}
SIGi(VSSi(τ))
end for

When a participant is ready to apply the update to their share, they perform the latter portion of Algorithm 10. That is, the participant verifies the update information received from other participants, and then applies the update to their share. The share update application phase is illustrated in Algorithm 14.

Algorithm 14 PUF-Share-Update-Application
Goal: Apply share update to construct ri(τ+1)
for all Participants pi ∈   do
Recover share ri(τ)
ri(τ) ← PUF-Retrieve(ci(τ), helperi(τ))
Verify ∀j ≠ i
DECi⁡(ENCi⁡(uji))·G⁢=?⁢∑k=1t⁢⁢ik⁢ϵjk
Update share as:
ri(τ+1)←ri(τ)+(∑j=1n⁢⁢uji⁢⁢mod⁢⁢q)
end for

As each threshold operation over a share can be performed independently of the other shares, the device need only recover one share at a time. This process is illustrated in Algorithm 15. Upon receiving a command and its associated auxiliary information Aux, the device first performs Algorithm 13 to prepare for the share update. Next, the device iteratively performs threshold operations over each share. A share is recovered by reading a challenge-helper pair from non-volatile memory, and using the PUF to regenerate the corresponding share. After performing a threshold operation over the share, the share update is applied using Algorithm 14, which generates the updated share for new time period (τ+1). After computing the threshold operations over each share, the threshold operations are combined to form the result which is returned to the server.

Algorithm 15 PUF-Threshold-OP-Staggered
Goal: Perform threshold operation
Assume: PUF-DKG (Algorithm 7) has been executed by PUF Device d
for Server s do
Issue Command     and Auxiliary Information Aux
end for
for PUF Device d do
for each share ri do
PUF-Share-Update-Preparation
end for
for each challenge ci (Challenge c = c0∥ . . . ∥cn) do
Read challenge ci and helper data hi
Recover share ri ←PUF-Retrieve (ci, hi)
Perform threshold operation    (ri, Aux)
Update share
ri(τ+1) ← PUF-Share-Update-Application(ri(τ))
Store ri(τ+1) and update PUF challenge:
{ci(τ+1) , helperi(τ+1)} ← PUF-Store(ri(τ+1))
end for
Combine threshold operations     ← Combine({   (ri, Aux)}0≤i≤n)
return Result
end for
for Server s do
Process operation
end for

In one embodiment, a (2,2) threshold system is constructed internally to the device. Algorithm 16 illustrates an example of a single-PUF (2,2) threshold construction of the more general Algorithm 15. The device has the share set {r₀, r₁}, and iteratively computes a threshold operation over each share to produce the set {,}. Once both threshold operations are complete and the shares have been updated and stored, the two threshold operations are combined into the final output .

Algorithm 16 Internal-PUF-Threshold-OP-Staggered
Goal: Perform threshold operation     at time τ
One-Time Setup Stage
for each Threshold Share ri do
Run Algorithm 6: Enrollment Publicize    ipub
end for
Run (2, 2)PUF-DKG Protocol, Publicize    pub
Evaluation Stage
for each Threshold Share ri do
PUF-Share-Update-Preparation
end for
for each Threshold Share ri do
Recover share ri(τ) ←PUF-Retrieve(ci(τ) , helperi(τ))
←    (ri(τ)), Local threshold operation
Update share
ri(τ+1) ← PUF-Share-Update-Application(ri(τ))
Store ri(τ+1) and update PUF challenge:
{ci(τ+1) , helperi(τ+1)} ← PUF-Store(ri(τ+1))
end for
return     ← Combine({   ,   })

The flow of Algorithm 16, a specific single-PUF (2,2) threshold construction of the more general Algorithm 15, is illustrated in FIG. 9. Prior to Step 1, the share update preparation (Algorithm 13) is performed. In Step 1, the first share r₀^τ is retrieved and its corresponding local threshold operation is performed. The share update (Algorithm 14) is then applied to r₀^τ to yield r₀^(τ+1) for the next time period. The updated share is then stored using a new random challenge c₀^(τ+1) which generates the corresponding helper data h₀^(τ+1) which will allow the updated share to be recovered using the PUF. The same process is followed in Step 2 for share r₁^τ. Finally, the combined output is constructed by combining the two local threshold operations that were performed over each share.

The device has a constant identity ^pub, ^priv, yet all operations that require ^priv are performed without ever reconstructing ^priv and with values that change after each operation is executed. As each part uses the PUF-Store and PUF-Retrieve algorithms to maintain their share, the (challenge, helper) pair is updated after each operation when PUF-Store is executed. Each share is refreshed for the new time period τ+1, and is stored by generating a new random challenge c_i^(τ+1) and setting the updated helper to helper_i^(τ+1)←ECC(r_i^(τ+1))⊕PUF(c_i^(τ+1)). Staggering the threshold operations such that the share regeneration, threshold operation, and share storing occur consecutively (rather than concurrently), precludes the simultaneous recovery of more than one updated share. Any tampering while one share exists would (assuming tampering pushes PUF output beyond error correction limits) prevent recovery of another share, in which case the device cannot perform operations over its private key.

An adversary applying a side channel attack against such an embodiment therefore must extract t or more shares from a period of observation that cannot exceed the period of refreshment. In other words, the adversary must compromise t devices in a given time period τ since any shares from time period τ are useless in time period τ+1. The difficulty of a side channel attack thus can be increased by updating more frequently (even after each operation). (Increasing refresh frequency also may multiply the difficulty inherent in side channel attacks on multiple-PUF device embodiments in which a remote adversary must observe and resolve multiple PUF values simultaneously generated in the device).

Also, whereas the longevity of systems using a fixed challenge/helper and response is directly limited to the hardware's increase in error rate due to aging, by continuously updating the pair in each time period, the error rate can be nominally reset to zero. That is, periodically refreshing the pair (c_i^(τ), helper_i^(τ)) during each time period τ links the PUF output to the current state of the hardware, eliminating the hardware drift, from previous time periods. In that regard, FIG. 11 illustrates the device recovering its share at time τ=1 using the original challenge-helper pair {c_i⁽⁰⁾, helper_i⁽⁰⁾} from time τ=0 using Algorithm 5: PUF-Retrieve. The device then internally generates a new challenge-helper pair {c_i⁽¹⁾, helper_i⁽¹⁾} for time period τ=1. The share is then stored by running Algorithm 4: PUF-Store using the new challenge-helper pair for τ=1. This links the updated challenge-helper pair to the current state of the hardware, which eliminates the hardware aging that occurred during time period τ∈[0,1). Thus, the expected number of bit errors in the PUF output at time τ=1 is zero despite the hardware continuing to age according to rate ρ.

As can be seen in FIG. 12, by repeating this process of periodically updating the internal challenge-helper pair of each PUF core, the maximum PUF output error can be bounded and made arbitrarily small by adjusting the refresh cycle period. Thus, a gradual shift in the PUF mapping is inconsequential. So long as the hardware has not fatally aged during the time between, the shift will be continuously factored into the stored challenge and helper data.

Dynamic Membership

The dynamic nature of shares in this construct also permits an embodiment in which the number of participants n participating in a group can be varied dynamically so that participants may join or leave the set of participants in the (t, n) threshold system. In this case, up to n−t participants can be removed from the set simply by leaving them out of the next share refresh protocol. To add a participant _j to the set of participants, each current participant _i generates an extra share u_ij from their share update polynomial δ_i(⋅).

In some embodiments employing dynamic membership (in a (t, n) threshold system) and multi-PUF device(s), the device(s) may be configured to perform a local self-test to ensure it is not nearing the point where it can no longer recover its shares due to hardware aging. A secondary threshold, t<i (the maximum number of errors that may be corrected by error correction), may be set for the device such that when t errors are observed in the PUF output a transition protocol is initiated. The transition protocol can transfer the ability of a device d_i to perform operations over its private key _i^priv to a different device d_j≠i without recovering _i^priv. In the example of a dual-PUF device, when device d_i detects critical hardware aging (e.g., when the PUF errors exceed secondary threshold t), it runs the share refreshing protocol and increases n: 2→4. Device d_i now possesses the set of shares {r_i}_1≤i≤4, and privately sends ENC_j{r_i}_3≤i≤4 to d_j after verifying that d_j is valid (e.g., verify the signature from a trusted source on d_j's enrollment token and have d_j perform a zero knowledge proof). Once d_j has received the set {r_i}_3≤i≤4, both d_i and d_j may act as d_i, and in the event of a hardware failure of d_i, it can be easily replaced by d_j.

The internal self-test procedure may be easily extended to the setting where multiple PUF-enabled devices are used as part of a larger system (see paragraph 74). When one PUF-enabled device fails to recover its share, it can be replaced with a new device. The remaining and correctly functioning PUF-enabled devices run the share update algorithm and increase n by sending the new device shares as well. This allows systems composed of multiple PUF-enabled devices to continue acting as a single entity, as failing devices can be immediately replaced and provisioned with shares of the global (t, n) threshold system.

Scalability

Standard PUF protocols are inherently linked to a specific hardware device (indeed, this is their goal), which can impose a constraint on the ability to readily scale a system to support an arbitrary processing load. FIG. 13 illustrates a processing center designed to scale to support an arbitrarily large processing load with PUFs employed to dynamically regenerate private information. By constructing a (t, n) sharing of a secret through distributed key generation, the private key for the system is never constructed or stored. However, any t PUFs can collaborate to perform cryptographic operations on behalf of the processing center. For example, if t=7, each row of PUFs can jointly perform a cryptographic operation on behalf of the processing center, and four requests (using rows A through D ) can be completed simultaneously. Thus, scalable hardware-intrinsic identity solutions can be designed in which a group of hardware components with local hardware-intrinsic identity (e.g., equipped with a PUF) are able to act cooperatively to form a unique hardware-intrinsic identity for their group as a whole. This embodiment of the present invention does not require the devices that compose the system to implement threshold cryptography locally. Rather, each device could run Algorithm 6 and publicize their local public key _i^pub. The (t,n) sharing is then set up for the system, using each device's local public key for private communication.

FIG. 14 illustrates a Master identity being generated from a set of component identities. An (n, n) threshold system could be constructed to require that all components are able to recover their share in order to perform cryptographic operations using the Master identity. In another embodiment of the invention, a (t,n) threshold system could be constructed to require that all of the critical components and some of the non-critical components are able to recover their share in order to perform cryptographic operations using the Master identity.

Performance

Performance tests were conducted on an embodiment having a single simulated 384-bit ring oscillator PUF (which did not require error correction), and a (2,2) threshold system defined over NIST elliptic curve P-384. Operations over each share were performed sequentially, rather than simultaneously as may be done in a multiple PUF embodiment. The tests measured the total time necessary for storage and retrieval of values, and communication between the device and the server. The server was equipped with an 8-core 3.1 GHz processor and 16 GB of RAM, and the device-side algorithms implemented on a Xilinx Artix 7 FPGA running at 100 MHz over a 115200 baud connection, where all operations were performed on the NIST P-384 curve. Table 1 reports the average time per protocol over 1000 trials.

TABLE 1
Performance Results
	Protocol	Average Runtime	St. Dev.
	Enrollment	0.011 seconds	0.0027
	Authentication	0.352 seconds	0.0002
	ElGamal Encryption	0.342 seconds	0.0014
	ElGamal Signature	0.352 seconds	0.0002

Claims

1. A device for securely executing cryptographic operations, the device comprising: a physical unclonable function (PUF) device;a memory; anda processor connected to the PUF device, the processor configured to: execute an arbitrary cryptographic operation in a sequence of ordered stages including at least a first stage associated with a first secret and a second stage associated with a second secret, the executing comprising: during the first stage: obtaining at least a first value from the PUF device;recovering the first secret based on at least the first value obtained from the PUF device;executing a first cryptographic operation using the first secret; andremoving from the memory one or more intermediate values generated by the act of recovering prior to execution of a subsequent stage;during the second stage and after the first stage: obtaining at least a second value from the PUF device;recovering the second secret based on at least the second value obtained from the PUF device; andexecuting a second cryptographic operation using the second secret and an output from the first cryptographic operation;execute the arbitrary cryptographic operation using an output of the second cryptographic operation; andprevent execution of the arbitrary cryptographic operation responsive to tampering with the device based on preventing recovery of at least one secret in the sequence of ordered stages.

2. The device of claim 1, wherein preventing execution of the arbitrary cryptographic operation responsive to tampering comprises preventing recovery of the second secret responsive to tampering prior to execution of the second stage.

3. The device of claim 1, wherein executing the arbitrary cryptographic operation comprises decrypting a data payload encrypted with at least the first secret and the second secret.

4. The device of claim 3, wherein: executing the first cryptographic operation using the first secret comprises executing a first decryption operation on the data payload using the first secret to obtain an intermediate output; andexecuting the second cryptographic operation using the second secret comprises executing a second decryption operation on the intermediate output using the second secret to decrypt the data payload.

5. The device of claim 1, wherein executing the arbitrary cryptographic operation comprises generating a digital signature with at least the first secret and the second secret.

6. The device of claim 5, wherein: executing the first cryptographic operation using the first secret comprises generating a first portion of the digital signature using the first secret; andexecuting the second cryptographic operation using the second secret comprises generating a second portion of the digital signature using the second secret.

7. The device of claim 1, wherein the processor is configured to: map the first secret to a first challenge using the PUF device; andmap the second secret to a second challenge using the PUF device.

8. The device of claim 7, wherein: the PUF device is configured to generate the unique values based on a challenge and the hardware characteristics of the PUF device; andthe processor is configured to: obtain the first value from the PUF device in response to transmitting the first challenge to the PUF device; andobtain the second value from the PUF device in response to transmitting the second challenge to the PUF device.

9. The device of claim 1, wherein removing the one or more intermediate values from the memory prior to execution of the subsequent stage comprises removing at least the first secret from the memory.

10. A method of securely executing cryptographic operations by a device, the method comprising: using a processor to perform: executing an arbitrary cryptographic operation, encoded with at least a first secret and a second secret, in a sequence of ordered stages including at least a first stage and a second stage, the executing comprising: during the first stage: obtaining a first value from a physical unclonable function (PUF) device;recovering the first secret based on at least the first value obtained from the PUF device;executing a first cryptographic operation using the first secret;removing from memory one or more intermediate values generated by the act of recovering prior to execution of a subsequent stage;during the second stage and after the first stage: obtaining a second value from the PUF device;recovering the second secret based on at least the second value obtained from the PUF device; andexecuting a second cryptographic operation using the second secret and an output from the first cryptographic operation;executing the arbitrary cryptographic operation using an output of the second cryptographic operation; andpreventing execution of the arbitrary cryptographic operation responsive to tampering with the device based on preventing recovery of at least one secret in the sequence of ordered stages.

11. The method of claim 10, wherein preventing execution of the arbitrary cryptographic operation responsive to tampering comprises preventing recovery of the second secret responsive to tampering prior to execution of the second stage.

12. The method of claim 10, wherein executing the arbitrary cryptographic operation comprises decrypting a data payload encrypted with at least the first secret and the second secret.

13. The method of claim 10, wherein executing the arbitrary cryptographic operation comprises generating a digital signature with at least the first secret and the second secret.

14. The method of claim 10, further comprising: obtaining the first value from the PUF device in response to transmitting a first challenge to the PUF device; andobtaining the second value from the PUF device in response to transmitting a second challenge to the PUF device.

15. The method of claim 10, wherein removing the one or more intermediate values from the memory prior to execution of the subsequent stage comprises removing at least the first secret from the memory.

16. A non-transitory computer-readable storage medium storing instructions that, when executed by a processor, cause the processor to perform a method comprising: executing an arbitrary cryptographic operation, encoded with at least a first secret and a second secret, in a sequence of ordered stages including at least a first stage and a second stage, the executing comprising: during the first stage: obtaining a first value from a physical unclonable function (PUF) device;recovering the first secret based on at least the first value obtained from the PUF device;executing a first cryptographic operation using the first secret; removing from memory one or more intermediate values generated by the act of recovering prior to execution of a subsequent stage;during the second stage and after the first stage: obtaining a second value from the PUF device;recovering the second secret based on at least the second value obtained from the PUF device; andexecuting a second cryptographic operation using the second secret and an output from the first cryptographic operation;executing the arbitrary cryptographic operation using an output of the second cryptographic operation; andpreventing execution of the arbitrary cryptographic operation responsive to tampering with the PUF device based on preventing recovery of at least one secret in the sequence of ordered stages.

17. The non-transitory computer-readable medium of claim 16, wherein preventing execution of the arbitrary cryptographic operation responsive to tampering comprises preventing recovery of the second secret responsive to tampering prior to execution of the second stage.