Authentication Prediction Dataset Using Click Patterns and Persuasive Cued Click Point
Description
This dataset contains 3000 records simulating user authentication attempts using click-based graphical passwords. It includes variables such as number of clicks, sequence length, success/failure rates, authentication methods (Draw-based, Pattern-based), and image category features (Animals, Grid, Landmarks, Letters, Nature, Numbers, Shapes, Symbols). The dataset supports research on authentication prediction using classical machine learning models and Persuasive Cued Click Point (PCCP) techniques.
Files
Steps to reproduce
User_ID: Unique identifier for each user. Num_Clicks: Total number of clicks during authentication. Sequence_Length: Length of the click sequence. Success: 1 if authentication succeeded, 0 otherwise. Failure: 1 if authentication failed, 0 otherwise. Click_Success_Ratio: Ratio of successful clicks to total clicks. Success_Rate: Overall success rate. Auth_Method_Draw-based, Auth_Method_Pattern-based: One-hot encoded authentication methods. Image_Category_*: One-hot encoded image categories (Animals, Grid, Landmarks, etc.). Split the dataset into training and testing sets (e.g., 80%-20%). Apply machine learning models: Logistic Regression Random Forest Gradient Boosting Support Vector Machine XGBoost CatBoost AdaBoost Extra Trees K-Nearest Neighbors Evaluate models using metrics: Accuracy AUC-ROC Precision Recall F1-Score Visualizations to Reproduce: Boxplot of features Feature correlation heatmap 3D scatter plot of Num_Clicks vs. Sequence_Length vs. Success Statistical Tests Performed: Chi-Square Test ANOVA Kruskal-Wallis Test Pearson Correlation Analysis Additional Details: Persuasive Cued Click Point (PCCP) method data is included and simulated for analysis. Deep Learning Model (simple ANN) also tested optionally.
Institutions
- Gandhi Institute of Engineering and Technology