Applying Machine Learning to Predict Symmetric Encryption Algorithm Inputs

The motivation for the topic of this thesis is to use Machine Learning to reverse engineer hash functions. Hash functions are supposed to be hard to reverse one-way functions. The Machine Learning algorithm will learn the hash function with a probability above 50 percent which means we can improve our guess of the inverse.This is done by implementing the DES symmetric encryption function to generate N many values of DES with a set key and the Machine Learning algorithm is taught a neural network to recognize the first bit of the input based on the value of the function's output. A new table is created and used for testing where the new table is created similarly but has different inputs. The Machine Learning algorithm, XGBoost using scikit-learn, runs on the new table and compares it to the other table and a confusion matrix is used to measure the quality of the guesses.