Cyberattack Prediction ML Model
Budget: €8 – €30 EUR
I have a labelled dataset that blends raw network-traffic captures, user-behaviour logs, packet-rate statistics, and detailed bulk/sub-flow metrics. Each record already includes a target column describing whether it is a DDoS, phishing, malware, brute-force, SQL-injection, or port-scan event.
The goal is to build a production-ready machine-learning pipeline that can (a) flag unseen anomalies and (b) classify any detected incident into one of the attack categories above. You are free to choose the most suitable approach—traditional algorithms, ensemble methods, or a deep-learning architecture—so long as the final model achieves robust, explainable performance on both tasks.
What I need is:
• Clean, well-documented code (Python preferred) that handles preprocessing, feature engineering, training, and evaluation
• A trained model file plus an easy way to retrain when new traffic arrives
• A concise report (or notebook) summarising metrics for classification accuracy, precision/recall per attack type, and anomaly-detection ROC-AUC
• Brief usage instructions so I can integrate the model into my existing security dashboard
Jupyter Notebook – full pipeline (load, clean, preprocess, train, evaluate)
Trained Models – .pkl files (main model + scaler + label encoder)
Results & Visuals – accuracy/F1 report, confusion matrix heatmap, feature importance plot
Project Report – 8–15 pages PDF (intro, method, results, discussion, conclusion)
Cleaned Dataset Sample – small CSV for reproducibility
Interactive Demo –
GitHub Repository – clean code, README, models, report, demo link
screen recording with steps
I need everything detailed every command and everything I should do, open or save on my computer fully explained.
The goal is to build a production-ready machine-learning pipeline that can (a) flag unseen anomalies and (b) classify any detected incident into one of the attack categories above. You are free to choose the most suitable approach—traditional algorithms, ensemble methods, or a deep-learning architecture—so long as the final model achieves robust, explainable performance on both tasks.
What I need is:
• Clean, well-documented code (Python preferred) that handles preprocessing, feature engineering, training, and evaluation
• A trained model file plus an easy way to retrain when new traffic arrives
• A concise report (or notebook) summarising metrics for classification accuracy, precision/recall per attack type, and anomaly-detection ROC-AUC
• Brief usage instructions so I can integrate the model into my existing security dashboard
Jupyter Notebook – full pipeline (load, clean, preprocess, train, evaluate)
Trained Models – .pkl files (main model + scaler + label encoder)
Results & Visuals – accuracy/F1 report, confusion matrix heatmap, feature importance plot
Project Report – 8–15 pages PDF (intro, method, results, discussion, conclusion)
Cleaned Dataset Sample – small CSV for reproducibility
Interactive Demo –
GitHub Repository – clean code, README, models, report, demo link
screen recording with steps
I need everything detailed every command and everything I should do, open or save on my computer fully explained.