1 Introduction

In 2001, National Institute of Standards and Technology (NIST) published AES as Federal Information Processing 197 (FIPS 197) [1]. AES was originally designed by Belgian researchers, Joan Daemen and Vincent Rijment. AES is a symmetric cryptographic algorithm which uses same secret key for both encryption and decryption operation. AES process data of block size 128 bits with different key sizes, 128,192, 256 bits. To get cipher text through AES algorithm plain text has to go through several rounds of AES process. Key size determines the number of rounds plain text has to be processed. AES algorithm with Key size 128 uses 10 rounds, 192 uses 12 rounds and 256 uses 14 rounds [2–6].

Due to the tremendous growth of technology, it is still possible to decode the 8 rounds in a 10 round AES. Remaining 2 rounds can also be brute forced with sufficient information on either plain text or secret key. Therefore to make AES as strong cryptographic algorithm it is necessary to modify the architecture of the existing algorithm. In this research work, the entire process in the encryption and decryption are made as secret key dependent. The following sections in this article is organized as follows Sect. 2 details the related research works of this research article, Sect. 3 explains the overview of the AES algorithm. In Sect. 4, proposed algorithm is explained in details, whereas in Sect. 5 has experimental results and its explanation and conclusion of this research is given in Sect. 6.

2 Related researches

Wadi et al. [7] given an approach to increase the speed of the execution of the AES algorithm by reducing the number of rounds of operation on plain text to obtain a single block The above Fig. 2 gives the simulation result of the proposed dynamic AES algorithm. Output is shown for two different keys with same plain text. Keys used for this simulation differ by a single bit and output can be measured for the avalanche effect analysis of cipher text.

Zhang et al. [8] proposed composite field arithmetic in s-box operation which increases the speed of the operation but also reduces the complexity of the s-box operation.

Rahimunnisa et al. [9] uses parallel processing and folded architecture to obtain high throughput.

Granado Criado et al. [10] designed an AES-128 algorithm with parallelism, pipelining, partial and dynamic reconfiguration. With new architecture a high th5roughput is achieved.

Jyrwa et al. [11] have worked with an iterative architecture and modifications including merging.

of SubBytes and ShiftRows, LUTs. This architecture yields throughput of 1.458 Gbps.

Sumio Morioka et al. [12] gives a modified S-Box circuit with low power PPRM architecture. This architecture reduces the power consumption by reducing the dynamic hazards and achieves lower consumption of 29 μW at 10 MHz.

Koraida et al. [13] described the performance of the modified AES algorithm by replacing the generalized matrix multiplication with the permutation table used in the DES process. With this modification, encryption and decryption becomes much faster than the existing algorithm.

Oyshee et al. [14] described an alternate approach for different s-box design with different Affine transformation process.

3 AES algorithm overview

In AES encryption and decryption operation the length of the input block, state and output block is 128 bits. Length of the secret key is 128,192 and 256 bits. Number of rounds of AES operation to be performed on a single block of data is determined by the key size. AES operations are performed as matrix calculation. Input plain text and subkey are transformed into matrix before being inserted into AES stages [15–19] (Table 1).

For both encryption and decryption rounds of operation AES uses four byte oriented transformations.

  1. 1.

    Sub byte transformation/inverse sub byte transformation

  2. 2.

    Shift rows/ inverse shift rows

  3. 3.

    Mix column/inverse mix column

  4. 4.

    Add round key

3.1 Sub byte transformation/inverse sub byte transformation

It is a non-linear byte to byte transformation technique, obtained as result of multiplicative inverse operation followed by affine transformation.

3.2 Shift rows

In this transformation, bytes in the state are shifted cyclically left shifted over different numbers for different rows. For first row, bytes are not rotated, where as second row shifted once, third row two times and fourth row is shifted three times byte by byte.

3.3 Mix column

This transformation technique is operated column by column in a state matrix. Output of the mix column technique is obtained after performing the multiplication of state matrix with the standard mix column matrix.

For Mix column operation

$$\left[ {\begin{array}{*{20}c} {S^{{\prime }} 1} \\ {S^{{\prime }} 2} \\ {S^{{\prime }} 3} \\ {S^{{\prime }} 4} \\ \end{array} } \right] = \left[ {\begin{array}{*{20}c} {02 03 01 01} \\ {01 02 03 01} \\ {01 01 02 03} \\ {03 01 01 02} \\ \end{array} } \right]\left[ {\begin{array}{*{20}c} {S1} \\ {S2} \\ {S3} \\ {S4} \\ \end{array} } \right]$$
(1)

For Inv Mix column operation

$$\left[ {\begin{array}{*{20}c} {S^{{\prime }} 1} \\ {S^{{\prime }} 2} \\ {S^{{\prime }} 3} \\ {S^{{\prime }} 4} \\ \end{array} } \right] = \left[ {\begin{array}{*{20}c} {0e 0b 0d 09} \\ {09 0e 0b 0d} \\ {0d 09 0e 0b} \\ {0b 0d 09 0e} \\ \end{array} } \right]\left[ {\begin{array}{*{20}c} {S1} \\ {S2} \\ {S3} \\ {S4} \\ \end{array} } \right]$$
(2)

3.4 Add round key

This is the only operation in which the secret gets operated with the pain text. A simple EXOR operation is carried out between the output of the mix column and subkey of that particular round.

3.5 Key scheduling

Key scheduling or Key expansion is a process of obtaining required number of subkeys one for each round of operation from the main secret key. Each subkey is dependent on the previous subkey. It’s a very important process, since there should not be any resemblance between any subkey. It involves word rotation, word shifting, addition of a constant value, EXOR operation [20–26].

4 Proposed AES technique

From the previous section, it is clear that the secret is used only in the add round key section. All the other sections are independent of secret key. Therefore those operations are static and can be easily revertible. To avoid this, a much more efficient AES algorithmic technique in which are the operations are performed based on the key. In shortly, all sections produce the result based on the key. Due to this property, operations on each round of proposed AES is dynamic.

4.1 Dynamic S-box transformation

S-box in the proposed work will not be same for all the rounds of AES operation. Therefore it is called as dynamic s-box depending on the key and subkey of that particular round of operation in AES. Because of this dynamic s-box, it is not easy for a third party attacker to crack the s-box without the information about the secret key.

In this work, S-box is made key dependent by the two following steps:

  1. (i)

    EXOR operation is performed between the byte values of the secret key or subkey of that particular round to find a resultant value.

  2. (ii)

    Based on the resultant value, the standard s-box value is rotated cyclically left. That rotated value is used as the replacement for the input value of the s-box stage.

For example: byte wise EXOR operation in the key b955b5ce8d20e346bcc2f146af68a5c1 gives F5. Therefore standard s-box values are rotated cyclically left F5 times to provide the final result. Input of 14 to the s-box stage gives 5F as the result in the proposed work instead of FA.

During decryption operation, s-box values are rotated cyclically right by the resultant value obtained as the result of byte wise EXOR operation in the subkey to give the desired output.

Pseudo-code for dynamic s-box transformation

figure a

4.2 Dynamic shift rows

In the standard Shift row technique, state matrix’s second, third and fourth rows are shifted cyclically left one, two, three times respectively. This technique is also static or same for the entire different types of subkey. In this proposed work, rows of the state matrix are shifted according subkey information. This technique is implemented in three steps.

  1. (i)

    Parity of the subkey of that particular round is identified by having the EXOR operation between all the bits of the subkey

  2. (ii)

    If the parity bit is ‘1′, second and fourth rows are shifted cyclically left by one and three times respectively.

  3. (iii)

    If the parity bit is ‘0′, second and third rows are shifted cyclically left by one and two times respectively.

Because of this property in this new technique, Shift row operation becomes dynamic based on the subkey parity value.

In decryption, the state matrix rows are rotated cyclically as same as in encryption, except rows are rotated right instead of left rotation.

4.3 Dynamic Mix column transformation

Like the previous stages of AES operation, In Mix column operation the matrix used to multiply the input is same for all rounds irrespective of the subkey. In this proposed work, a slight change is given to the mix column stage. Two matrixes are used to perform this operation. This process is implemented in three steps.

  1. (i)

    Parity value of the subkey of that particular round is calculated by applying EXOR operation between all the bits of the subkey

  2. (ii)

    If the parity value is ‘1′, the following matrix is used for mix column operation

    $$\left[ {\begin{array}{*{20}c} {S^{{\prime }} 1} \\ {S^{{\prime }} 2} \\ {S^{{\prime }} 3} \\ {S^{{\prime }} 4} \\ \end{array} } \right] = \left[ {\begin{array}{*{20}c} {02 03 01 01} \\ {01 02 03 01} \\ {01 01 02 03} \\ {03 01 01 02} \\ \end{array} } \right]\left[ {\begin{array}{*{20}c} {S1} \\ {S2} \\ {S3} \\ {S4} \\ \end{array} } \right]$$
    (3)
  3. (iii)

    If the parity value is ‘0′, the following matrix is used for mix column operation

    $$\left[ {\begin{array}{*{20}c} {S^{{\prime }} 1} \\ {S^{{\prime }} 2} \\ {S^{{\prime }} 3} \\ {S^{{\prime }} 4} \\ \end{array} } \right] = \left[ {\begin{array}{*{20}c} {0e 0b 0d 09} \\ {09 0e 0b 0d} \\ {0d 09 0e 0b} \\ {0b 0d 09 0e} \\ \end{array} } \right]\left[ {\begin{array}{*{20}c} {S1} \\ {S2} \\ {S3} \\ {S4} \\ \end{array} } \right]$$
    (4)

During decryption operation, matrix used in mix column gets interchanged for the generated parity value. For parity ‘1' Eq. 2 is used and for parity ‘0′ Eq. 1 is used for performing inverse mix column operation.

5 Experimental results

The above modified new dynamic AES algorithm is designed in verilog and verified its functionality using Modelsim and implemented in xilinx FPGA device xc6vlx75t-3ff784. Table 2 gives the simulation result of both standard AES algorithm and the proposed dynamic AES algorithm. Comparison between the cipher texts obtained on the simulation both the algorithms for a single bit change in the secret key is also provided in Table 3. Plain text is kept constant for the AES encryption operation. A single bit change is induced in the secret to measure the avalanche effect and check whether the new algorithm satisfies the strict avalanche criterion (Figs. 1, 2).

Table 1 AES Specification
Full size table
Table 2 Comparison of avalanche effect between standard AES and proposed dynamic AES
Full size table
Table 3 Comparison of avalanche effect
Full size table
Fig. 1
figure 1

Simulation result of proposed dynamic AES algorithm

Full size image
Fig. 2
figure 2

Comparison of avalanche effect

Full size image

5.1 Simulation result

5.1.1 Avalanche effect

For any cryptographic algorithm, the most desired property is Avalanche effect (4). It can be measured by changing a single bit in the input value and measuring the bits changes in the output for a single bit input change. Comparison between two outputs is made to determine the avalanche effect of the algorithm.

$${\text{Avalanche}}\;{\text{effect}} = {{n}}_{{\text{b}}} /{{t}}_{{\text{b}}}$$
(5)

where, nb = total number of bits changed in output. tb = total number of bits in output.

From the above Eq. 5 it is clear that the avalanche effect is directly proportional to the number bits changed in output due to a single bit change in input. Therefore high avalanche effect determines the large number of bits changed in output.

A cryptographic algorithm with high avalanche effect is harder for an attacker to crack the encryption process (5). It is wise to choose the algorithm which has high avalanche effect, to ensure that even a minute change has big impact on the output of the algorithm. For our proposed algorithm, 2000 samples of secret key are used for testing the avalanche effect. In this 1000 samples are differ in one bit on comparing with the other 1000 samples of key. Plain text remains same throughout the testing process. Because of the modification induced in this proposed algorithm, about 629 times the number of bits changed in the output is high in the proposed algorithm. About 265 times the number of bits changed in the output is high in the standard AES algorithm whereas, number of bit changes remains same in the both the algorithm is 106 times. Therefore our proposed algorithm is much more efficient since, it exhibits high avalanche effect when compared with the standard algorithm. Also from Table 2, it is clear that almost 50% of the bits get change in the output while employing proposed dynamic AES algorithm satisfying strict avalanche criterion. Table 3 and Fig. 2 gives the comparison result of avalanche effect between AES – 128 and our proposed work with 1000 pairs of samples are taken for the analysis.

5.1.2 Synthesis result

Above Table 4 gives the synthesis result of proposed algorithm. From the result it is clear that with sacrifice in the area, it is possible to achieve a logic circuit for which consumes low power as well as high avalanche effect.

Table 4 Synthesis report
Full size table

6 Conclusion

This article concentrates on improving the security of information in the UAV by implementing dynamic AES-128 algorithms. Different versions of AES have been designed are more efficient than the standard AES algorithm in terms of security other parameters. Some of the algorithms provide better performance in terms of area, power consumption and time complexity. With this newly proposed dynamic AES algorithm, the security of the cryptographic algorithm is enhanced by providing greater avalanche effect than the standard AES. By providing the key dependency in all the operations of AES, the avalanche effect has been increased in the output. About 58% increment has taken place in the avalanche effect in the proposed AES algorithm on comparing with the standard AES algorithm. This research article aims to improve only the avalanche effect on the output rather than concentrating area, time and power consumption by the proposed architecture.