Algorithm

Short Description Application Objective Function Performance Analysis Advantages Limitation Reference
Name
Deep Learning This approach has used back propagation DES To optimized deep Machine learning algorithms Better in - (Ankita
(MLP) technique with multiple hidden layers and Cryptanalysis neural network for (KNN) secured an accuracy accuracy, loss Mundra,
advanced activation function (LeakyReLU) cryptanalysis of range of 65–85% while this and time Shikhab
also we have addressed the problem of symmetric approach (Deep Learning) Mundra, Jai
vanishing gradient. The loss function used encryption algorithm secured an accuracy range of Shankera
was mean squared error. The optimizer used DES 86–91%. The losses and time Srivastava,
is Adaptive Moment Estimator (ADAM). taken during the training of the 2020)
plaintext and cipher-text were
also found much lower.

Deep Learning This paper proposes a generic cryptanalysis S-DES, To find the key of  When a text key is applied, This approach (So, 2020)
(MLP) model based on deep learning (DL), where Simon32/64, block ciphers from the DL-based attack broke limited the
the model tries to find the key of block and Speck32/64 known plaintext- the S-DES cipher with a keyspace to
ciphers from known plaintext-ciphertext cipher ciphertext pairs success probability of 0.9 only text-based
pairs. Activation function used is ReLU. To Cryptanalysis given 28.08 known plaintexts. keys
set the number of hidden layers and neuron  When a text key is applied,
used exhaustive searching. with a success probability of
0.99, the DL-based
cryptanalysis finds 56 bits of
Simon32/64 with 212.34
known plaintexts and 56 bits
of Speck32/64 with 212.33
known plaintexts.

Residual The vital step in differential cryptanalysis is Lightweight To evaluating the correct key was recovered with a Better results  Low (Tian and
Neural Network to find the differential characteristics of the Cipher SIMON safety of SIMON success rate of 23%, and the data than Accuracy Hu, 2021)
cryptographic primitives, so that an attacker cipher under the complexity and computation conventional  Round
can construct differential distinguishers to neural differential complexity are as low as 28 and differential Reduced
carry out the differential attack. cryptanalysis (key- 2 20.1 respectively cryptanalysis SIMON
recovery attack)

Rotational- Lightweight To apply the We construct related-key  Round (Koo, Jung

XOR Rectangle Cipher SIMON Rotational-XOR- rectangle distinguishers for Reduced and Kim,
Cryptanalysis difference (RXD) round-reduced versions of SIMON 2020)
approach to a Simon with block lengths of 32,  Does not
nonARX cipher 48, and 64, and we suggest a show the
Simon and evaluate five- or six-round key recovery best results
 its security attack for Simon

Nature Inspired Proposed method of cryptanalysis of S-DES S-DES  To study the  Cuckoo Search and Frefy (Kamal,
Optimization using nature inspired optimization cryptanalysis of Algorithms are superior than Bag and
Algorithms algorithms. The following experiments were S-DES via Genetic and Memetic Kule, 2021)
performed with an aim to outline the general nature-inspired Algorithms and can be used
performance, efectiveness and accuracy of meta-heuristic in diverse applications as per
CSA, FA and BHO Algorithm. algorithms the specifc requirements of
namely Cuckoo the problem while
Search, Firefy maintaining a tradeof
and Black-Hole between time and accuracy.
Optimization  If we consider this particular
Algorithms problem of cryptanalysis of
 To present a S-DES, the fve algorithms
comparative can be ranked as follows:
study on the CSA>FA>MA>BHO>GA.
efciency of these  CSA and FA are the most
three fairly new accurate out of all fve
algorithms with algorithms with an average
that of successful key bit recovery
previously of 80% and 84%
established respectively. GA, with 62%
Memetic average key bit recovery, is
Algorithm and shown to have the worst
Genetic performance while BHO and
Algorithms in MA have decent, close
regard to S-DES percentages of 72% and
cryptanalysis 73% respectively. The better
performances of CSA and
FA can again be attributed to
their sufciently randomized
nature strengthened by the
use of Lévy fights.
 REFERENCES

1. Ankita Mundra, Shikhab Mundra, Jai Shankera Srivastava, P. G. (2020) ‘Optimized deep neural network for cryptanalysis of DES’, Journal of
Intelligent & Fuzzy Systems, 38(5), pp. 5921–5931. doi: 10.3233/JIFS-179679.
2. Kamal, R., Bag, M. and Kule, M. (2021) ‘On the cryptanalysis of S-DES using nature inspired optimization algorithms’, Evolutionary
Intelligence, pp. 163–173. doi: 10.1007/s12065-020-00417-5.
3. Koo, B., Jung, Y. and Kim, W. H. (2020) ‘Rotational-XOR Rectangle Cryptanalysis on Round-Reduced Simon’, Security and Communication
Networks, 2020. doi: 10.1155/2020/5968584.
4. So, J. (2020) ‘Deep Learning-Based Cryptanalysis of Lightweight Block Ciphers’, Security and Communication Networks, 2020. doi:
10.1155/2020/3701067.
5. Tian, W. and Hu, B. (2021) ‘Deep learning assisted differential cryptanalysis for the lightweight cipher SIMON’, KSII Transactions on Internet
and Information Systems, 15(2), pp. 600–616. doi: 10.3837/tiis.2021.02.012.