A quantum encryption design featuring confusion, diffusion, and mode of operation

Quantum cryptography—the application of quantum information processing and quantum computing techniques to cryptography has been extensively investigated. Two major directions of quantum cryptography are quantum key distribution (QKD) and quantum encryption, with the former focusing on secure key distribution and the latter focusing on encryption using quantum algorithms. In contrast to the success of the QKD, the development of quantum encryption algorithms is limited to designs of mostly one-time pads (OTP) that are unsuitable for most communication needs. In this work we propose a non-OTP quantum encryption design utilizing a quantum state creation process to encrypt messages. As essentially a non-OTP quantum block cipher the method stands out against existing methods with the following features: 1. complex key-ciphertext relation (i.e. confusion) and complex plaintext-ciphertext relation (i.e. diffusion); 2. mode of operation design for practical encryption on multiple blocks. These features provide key reusability and protection against eavesdropping and standard cryptanalytic attacks.

www.nature.com/scientificreports/ practical secrecy 20 -i.e. breaking the encryption requires currently unrealistic computational resources. In this work we propose a new non-OTP quantum encryption design that utilizes a quantum state creation process to encrypt messages. Using a quantum state as the ciphertext, the quantum encryption offers an inherent level of protection against eavesdropping, because without the key any brute force measurement of the ciphertext state will collapse it into a random basis state. The non-readability of the ciphertext is a unique advantage of quantum encryption over classical methods where the ciphertext is just a bit string. Next we introduce the concepts of confusion (complex key-ciphertext relation) and diffusion (complex plaintext-ciphertext relation) from classical cryptography into quantum encryption and propose a novel encryption process that creates both confusion and diffusion. This ensures that small differences in the plaintext lead to substantial changes in the ciphertext or vice versa, such that the inability of a potential adversary to analyze the ciphertext state is amplified. Finally, we introduce the concept of mode of operation from classical cryptography into quantum encryption to enable practical encryption on arbitrary number of blocks of plaintexts. The mode of operation procedures developed for the quantum encryption design generalize the classical cipher block chaining (CBC) 23 to work with a quantum ciphertext by exploiting unique properties of quantum measurement and quantum superposition. The quantum mode of operation therefore has truly random or unreadable plaintext-altering materials that are impossible for the classical CBC mode. The adaptation of confusion, diffusion and mode of operation from classical cryptography into quantum cryptography not only provides key reusability and stronger security against standard cryptanalytic attacks but also establishes new design principles for the systematic development of quantum encryption methods which may lead to improved quantum cryptographic systems beyond the particular design of the current study.

Results
Encrypting classical data with quantum states. The essence of any encryption method with practical secrecy is a reversible process whose computational cost strongly depends on a secret piece of information called the key. In this work we focus on the symmetric-key scenario where decryption uses the same key as encryption. Consider an n-bit classical plaintext, practical secrecy is defined such that for the legitimate parties of the communication Alice and Bob knowing the key, both encryption and decryption are computationally simple in the sense that the number of computational steps required is polynomial: i.e. O cn k for some constant c and k such that cn k is overwhelmingly smaller than 2 n . In the meanwhile, for the adversary Eve not knowing the key, both encryption and decryption are computationally hard in the sense that the number of computational steps required is exponential: i.e. much greater than O(2 n ) . To achieve this with quantum encryption Alice starts with an n-qubit quantum state in the initial state |0� ⊗n . The first step Alice applies at most n Pauli-X gates to encode an n-bit classical plaintext into a quantum state plaintext: e.g. 00101 is coded into |00101� . The second step she applies a polynomial sequence of 1-qubit and 2-qubit elementary gates to transform the quantum plaintext into a quantum state that serves as the quantum ciphertext, and then sends it to Bob. The account of the polynomial sequence of elementary gates used by Alice is the key pre-shared with Bob such that upon receiving the quantum ciphertext Bob can apply the inverse operations to recover the quantum plaintext. The classical plaintext can then be revealed by projection measurement on the quantum plaintext in the computational basis. So far without going into any detail of the encryption procedure, the just described process is not so different from a generalization of existing studies of quantum encryption [17][18][19]24 , and we will later in "The quantum encryption with confusion and diffusion" and "Mode of operation" present the new quantum encryption design with confusion, diffusion, and mode of operation that provide key reusability and stronger security. However, here we first discuss certain security already provided by just considering the quantum nature of the ciphertext. Firstly, note the fact that a quantum state ciphertext naturally contains more uncertainty than a classical ciphertext. For example a classical bit 0 (1) can be mapped to a qubit state |0� ( |1� ), which after a unitary operation becomes a 1 |0� + a 2 |1� ( a * 2 |0� − a * 1 |1� ), where |a 1 | 2 + |a 2 | 2 = 1 . For encryption purpose a ciphertext in the form of a 1 |0� + a 2 |1� presents more difficulty to the eavesdropper Eve, because even if she has successfully intercepted the state a 1 |0� + a 2 |1� , without the key (i.e. the value of a 1 ) she cannot reliably read the content of the ciphertext. In practice if we assume a 1 can take N discrete values between 0 and 1, the uncertainty associated with it is typically far greater than 1 bit as N ≫ 2 . This difficulty for Eve is much more significant for a multi-qubit ciphertext state in which qubits are entangled with each other. This is because a brute-force measurement on the ciphertext state destroys the intricate dependencies among qubits and collapses the ciphertext into a simple state with all qubits in either |0� or |1� : such a state has little resemblance to either the ciphertext state or the plaintext. Consequently quantum encryption exploits the quantum phenomena of superposition and entanglement to produce a ciphertext that cannot even be read without the key. In comparison, a classical ciphertext is typically a bit-string with the same length as the plaintext, and it can be read and analyzed by Eve to gain information on the key and the plaintext.
Secondly, even if Eve is able to read the ciphertext-assuming the rare and can-be-avoided scenario that Alice sends the same ciphertext state many times and Eve is able to gain statistical knowledge of it-it is still highly difficult for her to deduce the key or the plaintext from the ciphertext. The detail of this reasoning is presented in the Supplementary Information S1 where the quantum state complexity theory in our previous study has been used 25 . Furthermore, this compromising scenario of Alice sending the same copy of the ciphertext many times can be totally avoided by the confusion, diffusion, and mode of operation to be introduced in the following sections.
The quantum encryption with confusion and diffusion. So far we have seen two security features by using a quantum state as the ciphertext: the difficulty in reading the quantum ciphertext and the impossibility to deduce the key even if the quantum ciphertext is somehow known. These features however are not sufficient for a www.nature.com/scientificreports/ good encryption method: to provide reusability of keys and protection against standard cryptanalytic attacks we need to design an encryption with good confusion and diffusion 20 . Confusion means complex relation between the ciphertext and the key such that it is difficult to deduce key properties by analyzing the patterns in ciphertexts. Classically if one bit in the ciphertext depends on multiple parts of the key, confusion is provided. For our quantum encryption design, as the ciphertext cannot be measured deterministically, confusion can be accordingly defined that the statistics of measuring one qubit in the ciphertext state depends on multiple parts of the key. Diffusion means complex relation between the plaintext and the ciphertext such that it is difficult to deduce plaintext properties by analyzing the patterns in ciphertexts or vice versa. Classically if changing one bit in the plaintext (ciphertext) changes more than half of the bits in the ciphertext (plaintext), diffusion is provided. Again since in our quantum encryption the ciphertext cannot be measured deterministically, diffusion can be defined that changing the value of one qubit in the plaintext leads to changes of statistics of measuring more than half of the qubits in the ciphertext. Note the vice versa ciphertext-to-plaintext relation is not defined for the quantum case because it is impossible to create a proper ciphertext without knowing the plaintext and the key first. We start with a basic encryption design where one unitary U i with real parameters (for simplicity we assume all parameters in the following discussions are real, however the method can be generalized to have complex parameters) is applied to each qubit q i of the plaintext, and no CNOT is applied. The key is then the collection {U i } where the order of U i 's is unimportant. Clearly this encryption does not provide either confusion or diffusion because the statistical pattern of measuring each qubit q i of the ciphertext depends on only one part of the key U i and only one qubit (the same q i ) of the plaintext. For example suppose after this step in the ciphertext q 1 = a 1 |0� 1 + a 2 |1� 1 and q 2 = b 1 |0� 2 + b 2 |1� 2 , then the probability of measuring |0� for q 1 is p(|0� 1 ) = a 2 1 and the probability of measuring |0� for q 2 is p(|0� 2 ) = b 2 1 . If this key is reused many times, Eve would be able to deduce U 1 and U 2 by measuring the probability of outcomes for q 1 and q 2 of the ciphertext (the same for all other qubits). Now after this step if we apply CNOT 1→2 (where 1 → 2 means q 1 is the control and q 2 is the target), the 2-qubit state is: -we see that q 2 gains a dependence on U 1 in the sense that the probabilities of outcomes when measuring q 2 depend on U 1 after CNOT 1→2 is applied. If we further apply CNOT 2→3 to q 3 = c 1 |0� 3 + c 2 |1� 3 , the 3-qubit state is: The results in Eqs. (1) and (2) reveal the effects of 1-qubit unitaries and CNOT's from a cryptographic perspective: Theorem 1 If the probabilities of outcomes when measuring a qubit depend on some 1-qubit unitaries applied to this or any other qubit, we say this qubit has dependences on these 1-qubit unitaries. Then a 1-qubit unitary creates dependences on its target qubit and a CNOT causes the target qubit to gain all the dependences from the control qubit, while the control qubit retaining all its dependences.

Proof of Theorem 1
Suppose q 1 is one qubit in a general n-qubit state φ (n) , the Schmidt decomposition of φ (n) with respect to q 1 is: where φ (n−1) 1 and φ (n−1) 2 are orthogonal, and therefore p(|0� 1 ) = C 2 1 a 2 1 + C 2 2 a 2 2 : this means q 1 depends on the pairs (C 1 , C 2 ) and (a 1 , a 2 ) that are created by previous quantum operations used to generate φ (n) . Now applying and thus the probabilities of no qubit other than q 1 are affected by U . Now suppose we further Schmidt-decompose φ = 0 , and then we can calculate the probability:p( 22 . We see that q 1 and q 2 share a dependence on the pair (C 1 , C 2 ) but the dependence on (a 1 , a 2 ) is unique to q 1 . Now apply CNOT 1→2 to φ (n) : where we see that q 2 has gained dependence on the pair (a 1 , a 2 ) , which was originally unique to q 1 . Because the form of φ (n) in Eq. (3) is entirely general, q 1 's dependence on (a 1 , a 2 ) can be understood as a package including all its dependences gained in the process of creating φ (n) -through either 1-qubit unitaries applied to q 1 or CNOT's applied to q 1 as the target. Equation (7) shows that by a single CNOT 1→2 all q 1 's dependences packaged in (a 1 , a 2 ) are created on q 2 . It is trivial to see that q 1 still retains its dependences. This concludes the proof for Theorem 1. Note that the dependences created on q 2 are not the same as those on q 1 -the probabilities indeed depend on the same unitaries, but the exact forms are different. Theorem 1 is significant that it allows us to create new probability dependences with 1-qubit unitaries on selective qubits and then efficiently pass them onto other qubits by CNOT gates. In the following we show how to use this result to design an encrypting process with good confusion and diffusion properties. The encrypting process with good confusion and diffusion: Start with an n-qubit plaintext where each qubit q i is either |0� or |1�.
Step 1: Apply a 1-qubit unitary U i to each qubit q i and create the initial dependence of each q i to its corresponding U i . This is the basic key design mentioned earlier. If each U i is defined by a real parameter that can take N discrete values, there are totally N n possibilities that contribute to key size. This step costs n U i gates.
Step 2: Apply CNOT i→i+1 sequentially for i = 1 to n − 1 : i.e. CNOT 1→2 first, then CNOT 2→3 , then CNOT 3→4 ,…, finally CNOT n−1→n . By Theorem 1, the CNOT 1→2 causes q 2 to gain the dependence on U 1 from q 1 , and then CNOT 2→3 causes q 3 to gain all the dependences from q 2 that include both U 2 from q 2 itself and U 1 that q 2 has just gained from q 1 . In such a snowball process, each further CNOT k→k+1 causes q k+1 to gain dependences on all the U i 's for i ≤ k . After this step each q i with i > n 2 (n even) or i > n+1 2 (n odd) has gained dependences on more than half of the U i 's. We remark that the order of the application of the CNOT i→i+1 gates is important: if we apply CNOT 2→3 before CNOT 1→2 , q 2 has not gained the dependence on U 1 from q 1 yet and thus q 3 will not gain that dependence either. Applying CNOT 2→3 before CNOT 1→2 is therefore less efficient than applying CNOT 2→3 after CNOT 1→2 as the latter can pass more dependences from q 2 to q 3 . This step costs n − 1 CNOT gates.
Step 3: When n is even, for each q i with i > n 2 (the downstream qubits), randomly assign a different q k with k ≤ n 2 (the upstream qubits), such that all the qubits are paired. When n is odd, disregard the n+1 2 th qubit and pair the remaining even number of (n − 1) qubits as just described. Apply CNOT i→k for each pair such that the upstream q k gains all the dependences from the downstream q i . After Step 2 each downstream q i with i > n 2 (n even) or i > n+1 2 (n odd) depends on more than half of the U i 's, and in Step 3 by the CNOT i→k gates these downstream qubits pass all their dependences to the corresponding upstream qubits. Consequently after Step 3 each one of the upstream qubits will have gained dependences on more than half of the U i ' , and this complex relation between the ciphertext and the key provides confusion as defined earlier. The process that gets all qubits into pairs has n 2 ! (n even) or n−1 2 ! (n odd) possibilities that contribute to key size. This step costs n 2 CNOT gates.
Step 4: Now to achieve diffusion defined earlier we want the property that changing the value of one qubit in the plaintext changes the statistics of measuring more than half of the qubits in the ciphertext. Suppose a qubit q j is |0� in the plaintext, after U j in Step 1 it becomes a 1 |0� j + a 2 |1� j and p(|0� 1 ) = a 2 1 . If the plaintext q j is changed to |1� then after U j it becomes a 2 |0� j − a 1 |1� j and p(|0� 1 ) = a 2 2 , so the dependence of q j on U j has changed. In addition, although the minus sign in a 2 |0� j − a 1 |1� j does not immediately have an effect on probabilities, it can change how the subsequent qubits depend on U j after Steps 2 and 3. Hence we see that a value change in one qubit q j in the plaintext will affect all the ciphertext qubits that have gained dependences from q j . This means that any upstream qubit q k with k ≤ n 2 already has diffusion after Steps 2, because all the downstream qubits in the ciphertext (more than half of all qubits) have gained dependences from q k . On the other hand the downstream qubits do not yet have diffusion after Step 2: e.g. no other qubit is dependent on q n because it is at the end of the chain of control in Step 2. Now to create diffusion in the downstream qubits, we just need to use these qubits as control and apply CNOT gates to random qubits as targets (can be either upstream or downstream) until on average more than half of all qubits have gained dependences from any qubit. For example, two qubits have gained dependences from the last qubit q n after Step 3: q n itself and the qubit assigned to pair with q n , thus we need to apply at most n 2 − 2 CNOT gates using q n as the control to pass q n 's dependences to half of all qubits. The actual CNOT gates required may be fewer than n 2 − 2 because we can first pass q n 's dependences to another downstream qubit such as q n−2 , and then any CNOT gate using q n−2 as the control will also pass q n 's dependences to the target. In fact, an example of a very efficient implementation is as shown in Step 4 of Fig. 1 to apply a series of CNOT gates running alternately through the downstream and upstream qubits, where the target qubit www.nature.com/scientificreports/ of the previous CNOT serves as the control qubit of the next CNOT: e.g. CNOT n→1 first, then CNOT 1→n−1 , then CNOT n−1→2 , then CNOT 2→n−2 , … , finally CNOT n / 2→n / 2+1 . By Theorem 1 it is easy to verify that this implementation guarantees more than half of all qubits have gained dependences from any downstream qubit. Unlike the previous steps, Step 4 allows greater freedom in the key design and the exact evaluation of the contribution to key size and gate cost is impossible. However, for the particular implementation just described, the order of the upstream qubits can be any permutation and thus there are n 2 ! possibilities that contribute to key size. This implementation costs n CNOT gates.
Step 4 concludes the ciphertext creation process. A graphical illustration of the four steps of encryption is drawn in Fig. 1. The account of all the unitaries and CNOT gates used is the key shared with the recipient, who can then recover the plaintext by reversing all the gates.
Through the description and analysis of the encrypting process, we can see that our quantum encryption design supports efficient implementation with O(n) gates and large key size with at least O N n n 2 ! possible variations. More importantly the design has provable confusion and diffusion that makes the key reusable while protecting against common cryptanalytic attacks. A worked-out 4-qubit example of the encryption process can be found in the Supplementary Information (S2).
Mode of operation. The quantum encryption described so far is a block cipher where each block of message containing n bits of classical information is encrypted into a quantum state of n qubits. Similar to the classical counterpart, the quantum block cipher also requires a mode of operation to ensure that different ciphertexts (blocks) are generated even with the same plaintext and key used. This feature together with diffusion allows the key to be reused many times to securely transmit large amount of information. Our mode of operation is inspired by the classical cipher block chaining (CBC) 23 . In the CBC mode a randomly chosen n-bit initialization vector (IV) is XORed ( ⊕ ) with the plaintext P 1 of the first block, the encrypting algorithm then works on IV ⊕ P 1 to produce the first ciphertext C 1 . Next C 1 is XORed with the plaintext P 2 of the second block before it is encrypted into C 2 . Repeat this process many times where each time the plaintext P i of the current block is XORed with the ciphertext C i−1 of the previous block before getting encrypted into the ciphertext C i of the current block: where E K () is the encrypting function with the key K. To generalize the CBC to our quantum encryption, the ciphertext here is a quantum state that cannot be directly XORed with the plaintext of the following block, and in the following we propose two different modes to solve this problem. Figure 1. Graphical illustration of the encrypting process with an 8-qubit example. The circles with numbers inside represent the qubits. The arrows represent CNOT gates for which each arrow begins at the control qubit and points to the target qubit. The numbers on the arrows indicate the order in which the CNOT gates are applied within the current step.
Step 1: apply a 1-qubit U i to each qubit q i . Step 2: apply CNOT i→i+1 sequentially for i = 1 to n − 1 , this step causes the downstream qubits 5-8 to gain dependences on more than half of the U i 's. Step 3: use the downstream qubits 5-8 as controls and the upstream qubits 1-4 as targets to apply CNOT gates. Showing one example out of the n 2 ! possible ways the qubits are paired. The CNOT gates in this step all commute so the order is unimportant. After this step confusion is achieved.
Step 4: with the general goal of achieving diffusion, this step has great freedom. In the particular example shown here, a series of CNOT gates run alternately between the downstream and upstream qubits. After this step diffusion is achieved. www.nature.com/scientificreports/ In the first mode shown in Fig. 2, after the first ciphertext state |C 1 � has been created |C 1 � = E K (P 1 ⊕ IV) , we create an additional copy of |C 1 � and measure it in the computational basis {|0�, |1�} . This will collapse the copy of |C 1 � into a classical bit string M(|C 1 �) , which can be then used to XOR with the plaintext of the following block to produce P 2 ⊕ M(|C 1 �) . We then encrypt this with E K (P 2 ⊕ M(|C 1 �)) = |C 2 � and send the recipient both M(|C 1 �) and |C 2 � . Repeat this process iteratively we have the general procedure: where M(|C i−1 �) is the measurement result on the extra copy of |C i−1 � . When the recipient has received M(|C i−1 �) and |C i � for each block after the first one, he decrypts with E −1 K (|C i �) = P i ⊕ M(|C i−1 �) , and then XOR with In the second mode shown in Fig. 3, after the first ciphertext state has been created by |C 1 � = E K (P 1 ⊕ IV) , we use the qubits of |C 1 � as controls to apply CNOT gates to the qubits of the following plaintext. Each qubit on |C 1 � as the control is paired with a different qubit on the following plaintext as the target. For simplicity, the same pairing plan that specifies which qubit of the current ciphertext state controls which target qubit of the next plaintext can be used for each iteration. Repeat this process iteratively: where CNOT(|C i−1 � → |P� i ) represents the altered plaintext after each qubit on the ciphertext state |C i−1 � as the control has applied a CNOT to a different qubit on the plaintext state |P� i as the target. When the recipient has received |C i−1 � and |C i � for each block after the first one, he decrypts with E −1 K (|C i �) = CNOT(|C i−1 � → |P� i ) to get the altered plaintext, and then use the qubits of |C i−1 � as controls to apply CNOT gates on the qubits of the altered plaintext to recover |P� i . Figure 2. The first mode of operation mechanism shown with a 3-block example. In each iteration after the first one, the extra copy of the ciphertext state |C i−1 � is measured into a classical bit string M(|C i−1 �) that is then XORed with the plaintext P i . www.nature.com/scientificreports/ Compared to the classical CBC, both quantum modes of operation have additional security because the material used to alter the plaintext for each iteration after the first one is not simply the ciphertext of the last block that is revealed to Eve. For the first mode, the bit string M(C i−1 ) for each iteration is generated with the truly random process of quantum measurement (as compared to pseudo-random number generation in classical computing) on the previous ciphertext state. For the second mode, all |C i−1 � 's are quantum states that cannot be reliably read. Furthermore, in the second mode the pairing plan of which qubit on the |C i−1 � controls which qubit on the next plaintext can be pre-shared as additional parts of the key-which has n! complexity. Both quantum modes of operation ensure different ciphertexts are generated even with the same plaintext and key used. Now comparing the two designs, the first mode is much easier to implement because each M(C i−1 ) as in Eq. (9) is a classical object and its XOR operation with the next plaintext is classical. On the other hand the second mode requires the ability to use the ciphertext state to control the next plaintext, which means more sophisticated quantum operations at both the encryption and the decryption ends. As a tradeoff the first design requires an additional classical channel to transmit the bit string M(C i−1 ) for each iteration (note this channel does not need to be secure because the bit string used to alter the plaintext in a mode of operation can be public without compromising security), while the second design only needs to pre-share two pieces of information: the initial IV and the pairing plan, and none other than the ciphertext is shared at the time of communication. Hence, the first design would be used when we prefer minimal quantum operations and have an additional non-secure classical channel available, while the second design would be used when we can afford more complex quantum operations and prefer to send a single ciphertext without additional channels. The increased key complexity through the pairing plan for the second design would also be a consideration.

Discussion
The mode of operation together with the encryption process completes our description of the new quantum encryption design. In actual application, Alice will first encode the classical bit string into a quantum basis state (e.g. 00101 is coded into |00101� ), and then apply a sequence of quantum gates following the procedure in "The quantum encryption with confusion and diffusion" to create a quantum ciphertext. Note that the procedure in "The quantum encryption with confusion and diffusion" is only a guideline to ensure confusion and diffusion by the result of Theorem 1. In this sense Theorem 1 can be considered as a foundational result that may inspire many other encryption procedures in addition to the particular one described in this work. Nonetheless the procedure in "The quantum encryption with confusion and diffusion" already provides great freedom with at least O N n n 2 ! variations contributing to the key size if a brute force attack is attempted. On the other hand the implementation cost of the procedure is only O(n) gates, which is very efficient. The ciphertext state can then be sent to Bob through an unsecure channel with possible eavesdropping by Eve. An account of the exact sequence of quantum gates applied by Alice is the key shared with Bob through a secure channel-note this can be done long before the actual communication happens thus it is harder to expect and attack by Eve. Upon receipt of the ciphertext state, Bob can apply the inverse quantum operations to recover the plaintext. After the first block of plaintext, additional blocks of plaintexts can be encrypted with additional mode of operation procedures as described in "Mode of operation" such that the statistics of the ciphertext state is further disguised.
The security of the quantum encryption design is provided by multiple mechanisms. Firstly the use of a quantum state as the ciphertext makes it impossible for Eve to reliably read and analyze the ciphertext. This is a unique quantum advantage over classical methods for which the ciphertext is just a bit string. In principle Eve could gain statistical knowledge of the ciphertext if the same one is sent many times, but this possibility is prevented by implementing one of the two quantum modes of operation. The two quantum modes of operation provide truly random or unreadable plaintext-altering materials depending on the mode of choice, and these are impossible for classical modes of operation. Having provable confusion and diffusion provides our method an additional layer of protection against potential cryptanalysis, because small changes in the plaintext lead to substantial changes in the ciphertext or vice versa. On the contrary, knowing the key, the legitimate recipient Bob can easily reverse the encrypting process to generate the plaintext deterministically from the ciphertext, without the need to actually read the ciphertext. The unique situation that the ciphertext can lead to the plaintext deterministically while not readable itself, together with features like confusion, diffusion, and mode of operation, make our quantum encryption strongly resistant to cryptanalytic attacks. For example, the chosen-plaintext attack (CPA) and the chosen-ciphertext attacks (CCA1 and CCA2) require Eve to analyze a few plaintext-ciphertext pairs to gain knowledge of the key. Now that the ciphertext being unreadable, and the statistics being obscured by confusion, diffusion, and mode of operation, it is very difficult for Eve to extract information from a few plaintext-ciphertext pairs. In addition, eavesdropping by Eve on the ciphertext inevitably disturbs the quantum state such that the recipient Bob can detect such interception. For Bob to determine if his measurement result is the correct message, the message disturbed by Eve, or the message corrupted by inherent system uncertainties (gate error, channel noise, etc.), multiple blocks of the same plaintext should be sent thus to establish a protocol analogous to the repetition code for error correcting purposes. As an interesting idea for future studies, the exact number of repetitions required for reliable communication should depend on the gate quality, channel quality, and key design.

Conclusion
In this work we have developed a quantum encryption design that utilizes a quantum state creation process to encrypt messages. By using a quantum state as the ciphertext and the creation procedure as the key, an inherent level of security is guaranteed by the statistical nature of quantum measurements as well as the complexity of the state creation process. We then introduce the concepts of confusion and diffusion from classical cryptography into quantum encryption and provide both features with a novel quantum encryption process. Finally we www.nature.com/scientificreports/ introduce the concept of mode of operation from classical cryptography into quantum encryption by proposing two modes of operation inspired by the classical CBC mode. The adaptation of confusion, diffusion and mode of operation from classical cryptography into quantum cryptography not only provides key reusability and stronger security against standard cryptanalytic attacks but also establishes new design principles for the systematic development of quantum encryption methods which may lead to improved quantum cryptographic systems beyond the particular design of the current study.