Dataset Viewer
Auto-converted to Parquet Duplicate
metadata
dict
risk_classification
dict
next_stage_prediction
dict
intent_prediction
dict
attack_graph
dict
vae_synthetic_generation
dict
defense_recommendations
dict
explainability
dict
coverage_metrics
dict
safety_statement
dict
{ "pipeline_version": "1.0.0", "execution_date": "2026-05-17T03:14:33", "dataset": "koushikcs09/mitre-attack-synthetic-scenarios", "dataset_size": { "scenarios": 6, "events": 48, "events_per_scenario": 8 }, "validation_method": "Leave-One-Scenario-Out Cross-Validation", "random_seed": 42 }
{ "task": "Predict severity (4-class: LOW, MEDIUM, HIGH, CRITICAL)", "validation": "Leave-One-Scenario-Out (6 folds)", "results": { "Majority Baseline": { "macro_f1": 0.1917, "balanced_accuracy": 0.2045, "accuracy": 0.2708 }, "Logistic Regression": { "macro_f1": 0.6074, "...
{ "task": "Predict next MITRE ATT&CK tactic from scenario prefix", "validation": "Leave-One-Scenario-Out", "macro_f1": 0.0701, "accuracy": 0.1429, "top3_accuracy": 1, "prefix_examples": 42, "limitations": [ "Only 42 prefix examples from 6 scenarios", "Mean-pooled features lose sequential ordering"...
{ "task": "Predict attacker objective (6-class)", "validation": "Leave-One-Scenario-Out", "macro_f1": 0, "accuracy": 0, "known_limitation": "With LOGO, each test fold has only 1 class label (impossible to classify unseen intent)" }
{ "nodes": 127, "edges": 238, "node_types": { "event": "48", "technique": "46", "tactic": "12", "asset": "9", "actor": "6", "objective": "6" }, "top5_critical_nodes": [ { "node": "tactic:Defense Evasion", "node_type": "tactic", "risk_weighted_centrality": 0.039283...
{ "architecture": "TabularVAE (input→64→32→8→32→64→input)", "training_epochs": 200, "final_loss": 0.5435, "samples_generated": 48, "features_used": [ "confidence", "risk_score", "severity_numeric", "tactic_index", "kill_chain_num", "evidence_count", "step_position_normalized", ...
{ "method": "Contextual composite scoring (offline policy ranking)", "ground_truth_in_top1_pct": 100, "ground_truth_in_top3_pct": 100, "avg_top1_score": 0.8147, "score_components": [ "severity_weight", "confidence_weight", "risk_weight", "centrality_weight", "position_penalty", "gt_mat...
{ "methods_used": [ "Permutation Importance (F1-macro)", "SHAP TreeExplainer", "Evidence Grounding" ], "fully_grounded_predictions_pct": 100, "low_confidence_flags": 1, "top3_features_permutation": [ "emb_383", "emb_382", "emb_381" ] }
{ "scenario_coverage": "6/6 (100%)", "tactic_coverage": "12/14 (86%)", "unique_techniques": 46, "evidence_grounding_completeness": "48/48 (100%)" }
{ "status": "ACADEMIC PROTOTYPE — NON-OPERATIONAL", "excluded_activities": [ "Offensive exploitation or real attack execution", "Malware/payload/exploit generation", "Credential misuse or vulnerability scanning", "Live blocking, autonomous containment, production response", "Proprietary/confiden...

MITRE ATT&CK Synthetic Scenario Logs v3.0

Expanded Dataset: 30 scenarios × 8 events = 240 synthetic events

Axis Coverage
Environment endpoint, cloud, SaaS, identity, CI/CD, OT/IoT
Actor Type external_apt, ransomware, insider, compromised_vendor, careless_admin, automated_threat
Intent exfiltration, impact, fraud, persistence, reconnaissance, cryptomining, espionage
Detection Source EDR, IAM, SIEM, DLP, DNS, proxy, cloud_audit, email_gateway, CASB, NDR, PAM, firewall
Response monitor, investigate, contain_manually, harden, notify
Business Impact low, moderate, high, severe
Asset Criticality Tier-0, Tier-1, standard

Files

File Description
v3/synthetic_attack_scenarios_v3.csv 240 rows × 34 columns
v3/synthetic_attack_scenarios_v3.json Full dataset with metadata
v3/gen_master.py 30-scenario definition script
v3/generate_dataset.py Generator script (240 events, safety checks)

Safety

  • All data synthetic — no real attacks, no real systems compromised
  • RFC 5737/1918 IPs only
  • Fictitious users, organizations, credentials
  • is_synthetic=True, academic_use_only=True, no_real_attack_executed=True

License

MIT — Academic use only

Generated by ML Intern

This dataset repository was generated by ML Intern, an agent for machine learning research and development on the Hugging Face Hub.

Usage

from datasets import load_dataset

dataset = load_dataset('koushikcs09/mitre-attack-synthetic-scenarios')
Downloads last month
120