text stringlengths 0 8.13M |
|---|
(Dated: October 28, 2021) |
We present authorized quantum computation, where only a user with a non-cloneable quantum |
authorization key can perform a unitary operation created by an authenticated programmer. The |
9002 securityofourauthorizedquantumcomputationisbasedonthequantumcomputationalcomplexity |
problemofforgingthekeysfromanobfuscatedquantumgatesequence. Undertheassumptionofthe |
existence of a sufficiently-random gate shuffling algorithm, the problem is shown to be in the NQP |
(Non-deterministic Quantum Polynomial)-hard class by reducing it to a NQP-Complete problem, |
the exact non-identity check problem. Therefore, our authorized quantum computation can be |
raM |
computationally secure against attacks using quantum computers. |
PACSnumbers: 03.67.Ac,03.67.Lx,03.67.Dd |
21 |
Considertheworldoncequantumcomputersexistand weshowasufficientconditionfortheobfuscationprocess |
]hp-tnauq[ arewidelyused. Inthisworld,unitaryoperationsarethe where forging the quantum authorization keys is compu- |
programsofquantumcomputers. Fortheprogrammerof tationally difficult even using quantum computers. We |
the quantum programs, they are important intellectual stressthatwedonotobfuscatethequantumprogramit- |
properties and it is important to protect the copyright self, but the identity of the quantum authorization keys. |
of the programs. On the other hand, for the user of the |
We note that if we do not require the authenticity |
program, they do not need to know about the details of |
of the program, blind quantum computation [3], which |
the program, they just want to perform a task, as long |
aims to perform a unitary operation without revealing |
as the created by the authenticated programmer. |
the identity of input states, can be used for similar |
1v8802.3090:viXra Such a situation can be solved if the programmer en- tasks. Blind quantum computation is a two-party proto- |
codes a program so that the original program is only colbasedoninformationalsecurityanditrequiresmulti- |
performable for users with non-cloneable authorization ple quantum/classical communications during computa- |
keys, distributes the encoded programs via an authen- tion. In contrast, our authorized quantum computation |
ticator, and sends the authorization keys directly to the is a semi-public protocol based on computational secu- |
users. Then,anyonecandownloadtheencodedprograms rity to ensure the authenticity of the program and no |
viatheauthenticator,whichareguaranteedtohavebeen communication is required during computation. |
made by the programmer, but it is performable only by |
We first present a construction of authorized quantum |
theauthorizeduser. Inthisletter,weproposeaschemeof |
computation to sketch our scheme. We regard unitary |
authorizedquantumcomputationthatallowsthistaskin |
operations to be quantum programs. Note that we con- |
quantumcomputationalsecurity,asapossiblenewquan- |
sider only polynomial quantum programs, namely, uni- |
tum cryptographic primitive. |
tary operations represented by an array of a polyno- |
In our scheme, we employ both quantum advantages mial number p(n) of elementary unitary gates. A uni- |
and quantum limitations for its security. To protect the tary operation U is described by a polynomial classi- |
originalprogramsfromunauthorizedusers, theencoding cal bit sequence {0,1}∗. We call this classical informa- |
process of the program is two-fold. One is an encryp- tion as a quantum gate sequence of U and denote it by |
tion process of introducing quantum authorization keys x(U). (Throughout this letter, we use capital letters to |
so that computation is not possible without using the represent unitary operations and small letters to repre- |
correct key. The keys are unknown quantum states for sent their quantum gate sequences.) Due to the non- |
anyusers(evenforauthorizedusers)andtheiranonymity uniqueness of the gate sequence representation for uni- |
and non-cloneability is ensured by quantum mechanics. tary operations, we consider a set of all quantum gate |
The other is an obfuscation process that hides the basis sequencesforaunitaryoperationU consistingofatmost |
of the keys in the encoded program. It is known that p(n) elementary gate arrays and denote it by g (U), |
p(n) |
classically, obfuscating programs is impossible [1]. For orsimplyg(U)ifthespecificationofafunctionp(n)isir- |
quantum settings, existence of obfuscation with the help relevant. Since calculations of the matrix representation |
of quantum states is an open problem [2]. In this letter, ofaunitaryoperationrequireexponentialcomputational |
by introducing the concept of gate shuffling algorithms, power in terms of n, we have to rely on the polynomial |
2 |
gate sequence representation of unitary operations when |
n is large. Now consider a programmer who wants to |
encode a unitary operation U (where i is an index for |
i |
specifying the unitary operation) acting on a n-qubit in- |
put Hilbert space H⊗n . |
input |
Step 1: The programmer extends the U into another |
i |
unitary operation G acting on a larger Hilbert space |
H⊗(m+n) by adding a m-qubit Hilbert space H⊗m of |
key |
quantum authorization keys in front of the input Hilbert FIG. 1: A construction of gate sequences for the unitary op- |
space. (We often simply denote the quantum authoriza- erations G and G(cid:48). See the text for notations. |
tion key as the key.) This extension is similar to the |
programmablequantumgatearraysproposedbyNielsen |
desired state and is highly likely to be entangled. The |
andChuang[4]. TheextendedunitaryoperatorGtrans- |
security of our scheme will be discussed later. |
forms G(|i(cid:105)⊗|ϕ(cid:105)) = |i(cid:105)⊗U |ϕ(cid:105), where {U } is a set of |
i i |
P =2k (1≤i≤2k (cid:28)2m) unitary operations for an ar- We note that by performing the reverse quantum gate |
bitrary input state |ϕ(cid:105)∈H⊗n , and {|i(cid:105)∈H⊗m} is the sequenceusingtheusedkeystate|φ(cid:48) i(cid:105),thereverseunitary |
corresponding key states inin apu cot mputational bk ae sy is speci- operation U† can be also performed and the original key |
i |
fied by a binary bit sequence {0,1}k. The number of the |φ (cid:105)isregainedbyG(cid:48)†(|φ(cid:48)(cid:105)⊗|ϕ(cid:105))=|φ (cid:105)⊗U†|ϕ(cid:105). There- |
i i i i |
key qubits k should be taken to be of order log(n) for fore, we can recycle the quantum authorization key as |
restricting the total gate number of G to be in polyno- long as it keeps coherence. |
mialofn. Thusthe(m−k)-qubitdummyspaceH⊗(m−k) In this scheme, security depends on the obfuscation |
dummy |
is introduced in the key space. A construction of G for process given in Step 3. To present a sufficient condition |
{U } is shown in Fig. 1, where M and M are random for obfuscation, we investigate strategies for a malicious |
i 1 2 |
unitary operations acting on the dummy qubit space. user, Eve. We consider that Eve wants to perform the |
Step 2: By applying random unitaryoperations L and originalunitaryoperationU i withoutusingthequantum |
R on only H⊗m as G → G(cid:48) = (L⊗I)G(R⊗I) where I authorizationkey|φ i(cid:105)issuedbytheprogrammer. Apow- |
key |
denotesanidentityoperatorofanappropriatedimension, erful Eve may also be able to tap the quantum channels |
we create a key state |φ i(cid:105)=R†|i(cid:105) satisfying and steal other N keys {|φ j(cid:105)} j=1,...,N where j (cid:54)= i, and |
analyze the quantum gate sequence x(cid:48)(G(cid:48)) to obtain a |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.