Welcome to the Avengers Ensemble repository, home to cutting-edge authorship obfuscation techniques using MUTANT-X. Elevate your text security with our implementation that surpasses the capabilities of existing methods.
Dive into the realm of Mutant-X authorship obfuscation, inspired by the groundbreaking research paper, "Mutant-X: Efficient Authorship Obfuscation Techniques." Our approach refines the original concept by introducing the Ensemble Vote classifier and harnessing the power of writeprint features to enhance SVM classifiers' training.
- Robust ensemble attribution classifier with a 10-SVM classifier pipeline.
- Ensemble Vote classifier dynamically optimizes the ensemble by excluding low-accuracy classifiers.
- Writeprint features, including:
- Average character per word
- Letter frequency
- Key letter bigrams and trigrams
- Digits and characters percentages
- Frequency of digits and word lengths
- Integration of Word2vec model (gensim) for efficient word neighbor computations.
- MUTANT-X classifier with SVC, trained on the Amazon review dataset as an internal attribution classifier.
- Local dictionary for storing computed neighbors, enhancing computation efficiency.
- Achieved an outstanding average METEOR score of 0.5 on the Amazon review dataset, surpassing the implemented paper's results.
Empower your projects with Mutant-X authorship obfuscation:
- Clone Repository: Copy this GitHub repository to your local machine.
- Install Dependencies: Install the necessary dependencies listed in the requirements.txt file.
- Prepare Input: Create the input text file you wish to obfuscate.
- Run Script: Execute the main.py script with the input text file as a command-line argument.
- Obfuscated Output: Retrieve the obfuscated text, saved in a file with the original name appended by "_obfuscated."
Revolutionize your text security with Mutant-X authorship obfuscation. Our project excels by combining Ensemble Vote classification, advanced writeprint features, and a meticulously trained MUTANT-X classifier. Notably, the local dictionary optimizes computation time, making the obfuscation process swift and effective. Experience superior results on the Amazon review dataset, marking a significant improvement over the original paper's implementation. Join us in shaping the future of authorship obfuscation!