•  
  •  
 

Abstract

An important part of a cryptosystem is a cryptographic algorithm, which protects unauthorizedattackers from obtaining private and sensitive data. This study is a research project on identifying cryptographicalgorithms using deep learning techniques and categorizing cryptographic algorithms based on feature extraction. Theresearch involves employing block cipher modes called electronic codebook with the encryption algorithms Blowfishand advanced encryption standard (AES), where the data will be encrypted using the same key and a different key.The model has been developed by changing the structure and parameters of the proposed model and the training rateof the data. This model will build several dense FCNN of n layers on regular fully connected neural networks. Itsconstruction will consist of five hidden layers, with each layer consisting of 128 neurons and hidden layers activationRelu except for the output layer, which consists of two classifiers and the SoftMax activation function. FCNN isbetter able to classify big data. It is also more efficient in use, reducing complexity, with the ability to store trainingdata.First, the fully connected neural network (FCNN) model was used to evaluate the categorization of the models.Then, all models, even the encryption forms, were evaluated utilizing true positive measurements for satisfactoryclassification of the identified encryption method and false positive measurements for incorrect classification. Theeffectiveness of the model was then calculated using the precision value, recall, loss, accuracy range, and F1-Scoremetrics using a confusion matrix. The FCNN model parameters will be changed to more effectively identify theencryption algorithm. In the proposed method, when using the same key, the accuracy was 81%, and when using adifferent key, the accuracy was 49%. The FCNN model’s adjusted weights and learning will be based on large datato define and assess encryption algorithms more effectively and efficiently

Download

Share

COinS