A Dynamic nDES Model for Hiding Datasets for Machine Learning

The Data Encryption Standard (DES) is a block cipher for converting plain text into ciphertext. It is a symmetric key algorithm grounded on the Feistel network. The DES uses 56-bit keys to complete encryption in 16 rounds. However, on its own, the DES alone maybe brute force attacked. Using the DES model multiple times (number of times) may strengthen the key size. Improvement of the DES model for hiding the datasets used for machine learning purposes is the theme of this paper. A depth parameter is introduced which triggers the generation of hard to predict sequencing, giving rise to the nDES model. We investigate successful application of the depth parameter, and the sequencing approach in an effort to improve the DES model for hiding datasets. Results indicate that depth parameters between 1 and 8 are commonly influenced by the DES’s key size of 8 characters. In addition, the selected depth determines the complexity of the sequencing thereto. However, the depth does not influence the CPU time required to execute the proposed nDES model. Hence, these are inexpensive improvements to the traditional DES model.