generate_dataset.py

import pandas as pd

def generate_variations(base_name):
    words = base_name.split("_")
    variations = set()

    # Original case
    variations.add(base_name)

    # Snake case (lower & upper)
    variations.add("_".join(words).lower())
    variations.add("_".join(words).upper())

    # Pascal case
    variations.add("".join(word.capitalize() for word in words))

    # Camel case
    variations.add(words[0].lower() + "".join(word.capitalize() for word in words[1:]))

    # Kebab case
    variations.add("-".join(words).lower())

    # No separators (concatenated)
    variations.add("".join(words).lower())
    variations.add("".join(words).upper())

    return variations

# Base field names (HIPAA-sensitive)
hipaa_field_names = [
    # Name Variations
    "first_name", "last_name", "middle_name", "full_name", "given_name", "surname", "alias_name",
    "nick_name", "initials", "maiden_name", "patient_name", "legal_name", "preferred_name",
    "registered_name", "official_name",

    # Date of Birth Variations
    "dob", "date_of_birth", "birthdate", "birth_date", "DATEOFBIRTH", "Birth_Age", "birth_year",
    "birth_month", "date_of_birth_year", "DOB_YMD", "dob_day", "date_of_birth_full", "year_of_birth",
    "day_of_birth", "birth_details",

    # SSN Variations
    "ssn", "social_security_number", "SSN_HASHED", "encrypted_ssn", "ssn_number", "SSN_ENCRYPTED",
    "ssn_last4", "ssn_partial", "masked_ssn", "national_ssn",

    # Address Variations
    "home_address", "residential_address", "mailing_address", "current_address", "permanent_address",
    "street_address", "house_address", "physical_address", "registered_address", "patient_address",
    "billing_address", "delivery_address", "home_location", "residence_location",

    # Phone Variations
    "phone_number", "home_phone", "work_phone", "mobile_number", "cell_phone", "contact_number",
    "patient_phone", "emergency_phone", "alternate_phone", "fax_number", "guardian_phone",

    # Email Variations
    "email_address", "personal_email", "work_email", "user_email", "primary_email", "email_id",
    "contact_email", "patient_email", "guardian_email", "alternate_email", "notification_email",

    # Patient ID Variations
    "patient_id", "patient_identifier", "patient_number", "patient_uid", "patient_record_id",
    "medical_record_id", "health_record_id", "patient_reference", "individual_patient_id",

    # Medical Record Number Variations
    "medical_record_number", "MRN", "record_number", "health_record_id", "patient_record_number",
    "hospital_record_number", "case_file_number", "admission_record_id", "clinic_record_id",

    # Insurance ID Variations
    "insurance_id", "policy_number", "health_insurance_number", "medical_policy_id", "insurance_plan_id",
    "payer_policy_number", "coverage_id", "insurance_card_number", "group_policy_id", "subscriber_id",

    # Credit Card Variations
    "credit_card_number", "debit_card_number", "card_number", "masked_card_number", "encrypted_card_number",
    "hashed_card_number", "tokenized_card_number", "card_expiration_date", "card_expiry", "card_expiry_month",
    "card_expiry_year", "card_cvv", "cvv_code", "cvv2", "card_security_code", "card_verification_value",
    "cardholder_name", "cardholder_address", "billing_name", "billing_address", "billing_street", "billing_city",
    "billing_state", "billing_zip", "billing_postal_code", "billing_country", "billing_phone", "billing_email",
    "payer_name", "payer_email", "payer_id", "payment_amount", "payment_currency", "payment_date", "payment_status",
    "payment_reference", "payment_token", "payment_authorization", "payment_processor", "payment_gateway",
    "payment_method", "transaction_id", "transaction_reference", "transaction_status", "transaction_amount",
    "transaction_currency", "transaction_date", "transaction_auth_code", "transaction_settlement_date",
    "transaction_fees", "merchant_id", "merchant_account_id", "merchant_reference", "charge_id", "refund_id",
    "settlement_id", "reversal_id", "dispute_id", "invoice_id", "receipt_id", "order_id", "subscription_id",
    "recurring_payment_id", "statement_descriptor", "authorization_code", "authorization_expiry", "tokenization_status",
    "fraud_detection_status", "risk_assessment_score", "stripe_card_number", "stripe_card_token",
    "stripe_card_fingerprint", "stripe_card_last4", "stripe_card_expiration", "stripe_card_expiry_month",
    "stripe_card_expiry_year", "stripe_card_brand", "stripe_card_country", "stripe_payment_method",
    "stripe_payment_intent", "stripe_payment_status", "stripe_payment_currency", "stripe_payment_amount",
    "stripe_transaction_id", "stripe_charge_id", "stripe_customer_id", "stripe_session_id", "stripe_billing_details",
    "stripe_invoice_id", "stripe_refund_id", "stripe_balance_transaction", "stripe_dispute_id", "stripe_setup_intent",
    "stripe_subscription_id", "stripe_subscription_status", "stripe_subscription_start_date",
    "stripe_subscription_end_date", "stripe_subscription_interval", "stripe_subscription_plan",
    "stripe_subscription_trial_end", "stripe_payout_id", "paypal_email", "paypal_payer_id", "paypal_payer_status",
    "paypal_transaction_id", "paypal_transaction_status", "paypal_payment_id", "paypal_payment_token",
    "paypal_payment_status", "paypal_payment_currency", "paypal_payment_amount", "paypal_payment_date",
    "paypal_payment_reference", "paypal_billing_agreement", "paypal_billing_plan_id", "paypal_order_id",
    "paypal_invoice_id", "paypal_subscription_id", "paypal_subscription_plan", "paypal_subscription_status",
    "paypal_subscription_start_date", "paypal_subscription_end_date", "paypal_subscription_interval",
    "paypal_refund_id", "paypal_merchant_id", "paypal_settlement_id", "paypal_reversal_id", "paypal_dispute_id",
    "paypal_payout_batch_id", "paypal_payout_id", "braintree_customer_id", "braintree_payment_method",
    "braintree_transaction_id", "braintree_transaction_status", "braintree_transaction_reference",
    "braintree_card_number", "braintree_card_last4", "braintree_card_expiration", "braintree_card_expiry_month",
    "braintree_card_expiry_year", "braintree_card_token", "braintree_payment_nonce", "braintree_payment_token",
    "braintree_payment_status", "braintree_payment_currency", "braintree_payment_amount", "braintree_subscription_id",
    "braintree_subscription_status", "braintree_subscription_start_date", "braintree_subscription_end_date",
    "braintree_subscription_plan", "braintree_subscription_interval", "braintree_merchant_account_id",
    "braintree_settlement_batch_id", "braintree_refund_id", "braintree_dispute_id", "square_card_nonce",
    "square_payment_id", "square_transaction_id", "square_transaction_status", "square_transaction_reference",
    "square_payment_status", "square_payment_currency", "square_payment_amount", "square_payment_date",
    "square_customer_id", "square_order_id", "square_invoice_id", "square_subscription_id", "square_subscription_plan",
    "square_subscription_status", "square_payout_id", "square_refund_id", "square_dispute_id",
    "adyen_payment_reference", "adyen_transaction_id", "adyen_transaction_status", "adyen_payment_status",
    "adyen_payment_currency", "adyen_payment_amount", "adyen_payment_date", "adyen_card_alias", "adyen_card_expiry",
    "adyen_card_last4", "adyen_card_brand", "adyen_psp_reference", "adyen_merchant_account", "adyen_settlement_id",
    "adyen_refund_id", "adyen_dispute_id", "authorize_net_transaction_id", "authorize_net_transaction_status",
    "authorize_net_payment_status", "authorize_net_payment_currency", "authorize_net_payment_amount",
    "authorize_net_payment_date", "authorize_net_card_number", "authorize_net_card_last4", "authorize_net_card_expiry",
    "authorize_net_refund_id", "authorize_net_settlement_id", "authorize_net_dispute_id", "worldpay_transaction_id",
    "worldpay_transaction_status", "worldpay_payment_status", "worldpay_payment_currency", "worldpay_payment_amount",
    "worldpay_payment_date", "worldpay_card_number", "worldpay_card_last4", "worldpay_card_expiry",
    "worldpay_payment_token", "worldpay_settlement_id", "worldpay_refund_id", "worldpay_dispute_id"

    # Banking Variations                                                                    "bank_account_number", "account_number",
                                                                              
    "routing_number", "bank_details", "iban_number",
    "account_holder_name", "financial_account_number", "masked_account_number", "bank_identifier",

    # Diagnosis Variations
    "diagnosis_code", "diagnosis_description", "medical_condition", "patient_diagnosis", "health_diagnosis",
    "clinical_diagnosis", "diagnostic_code", "disease_code", "diagnosis_notes", "medical_findings",

    # Treatment Plan Variations
    "treatment_plan", "medical_treatment", "therapy_plan", "treatment_procedure", "treatment_guideline",
    "prescribed_treatment", "patient_treatment", "doctor_recommendation", "care_plan", "healthcare_plan",
    "allergy"

    # Blood Type Variations
    "blood_type", "blood_group", "patient_blood_type", "blood_classification", "ABO_group", "RH_factor",

    # Security & Access Control Variations
    "user_id", "login_id", "token", "security", "access_token", "session_token", "audit_log_id", "access_timestamp",
    "access_reason", "security_event_id", "breach_notification_status", "multi_factor_auth_enabled",
    "account_locked", "password", "password_hash", "encryption_status", "HIPAA_compliance_status",
    "authentication_token",
    "system_user_id", "access_logs", "authorization_token", "user_role", "security_question", "jwt",
    "username", "email", "password_salt", "encrypted_password", "hashed_password", "refresh_token", "session_token",
    "api_key", "secret_key",
    "oauth_token", "jwt_token", "auth_token", "mfa_token", "otp_code", "otp_secret", "two_factor_code", "multi_factor",
    "multi_factor_auth_enabled", "account_status", "failed_login_attempts", "login_attempts", "last_login",
    "login_timestamp", "security_answer",
    "recovery_email", "recovery_phone", "password_reset_token", "password_reset_expiration", "biometric_hash",
    "fingerprint_hash", "face_id", "iris_scan", "voice_auth",

]

# Base field names (Non-HIPAA-sensitive)
non_hipaa_field_names = [
    "record_id", "created_at", "updated_at", "modified_at", "deleted_at", "status", "active_status", "category", "tag","notes",
    "comments", "description", "reference_code", "version", "approval_status", "approval_date", "document_type","file_name",
    "file_path", "file_size", "file_extension", "attachment", "metadata", "tags", "priority", "rank", "rating","feedback", "flag",
    "is_active", "is_deleted", "is_verified", "workflow_status",
    "pipeline_stage", "job_status", "error_code", "error_message",
    "response_status", "response_code", "operation_mode", "settings", "preferences", "timestamp",
    "created_by", "modified_by", "approved_by", "verified_by",

    # Product & Inventory Fields
    "product_id", "product_name", "product_category", "product_description", "product_sku", "product_barcode",
    "product_price", "product_quantity", "stock_level", "warehouse_location", "supplier_name", "supplier_id",
    "supply_chain_status", "shipment_status", "order_status", "inventory_count", "purchase_order_id",
    "reorder_level", "restock_date", "delivery_date", "shipment_tracking", "batch_number",
    "manufacturing_date", "expiry_date", "unit_price", "discount", "tax_rate", "tax_amount", "profit_margin",
    "cost_of_goods_sold", "sales_count", "return_rate", "refund_status", "refund_reason",

    # System & Logging Fields
    "log_id", "event_type", "event_details", "log_level", "log_timestamp", "log_source", "log_category",
    "log_message", "event_status", "action_performed", "error_logs", "warning_logs", "debug_message",
    "performance_metrics", "resource_usage", "cpu_usage", "memory_usage", "disk_space", "execution_time",
    "thread_id", "process_id", "request_id", "session_duration", "system_uptime", "server_name",
    "server_region", "api_endpoint", "request_method", "response_time", "status_code",

    # Geographic & Location Data
    "city_name", "state_name", "region", "country", "continent", "time_zone", "latitude", "longitude",
    "altitude", "geolocation_code", "weather_conditions", "climate_zone", "temperature", "humidity",
    "wind_speed", "air_quality", "location_id", "map_reference", "landmark", "route_name", "road_condition",

    # Organizational & HR Fields
    "department_id", "department_name", "team_name", "team_lead", "employee_id", "position", "role",
    "designation", "work_shift", "office_location", "branch_name", "branch_code", "payroll_status",
    "payroll_cycle", "leave_status", "attendance_status", "work_hours", "project_name", "project_id",
    "task_id", "task_status", "milestone", "project_budget", "operating_expense", "revenue_stream",
    "business_unit", "performance_rating", "bonus_eligibility", "promotion_status", "training_completion",

    # Finance & Accounting Fields
    "account_type", "financial_year", "budget", "expense_category", "expense_amount", "fund_source",
    "ledger_code", "revenue_growth", "operating_income", "profit_loss", "investment_type",
    "stock_market_trend", "portfolio_value", "dividend_yield", "asset_id", "asset_category",
    "asset_lifetime", "depreciation", "return_on_investment", "cost_savings", "payment_mode",
    "billing_cycle", "pricing_strategy", "profit_forecast", "economic_indicator",

    # Marketing & User Engagement Fields
    "campaign_id", "campaign_name", "ad_group", "advertisement_id", "conversion_rate",
    "click_through_rate", "impressions", "page_views", "email_open_rate", "bounce_rate",
    "user_feedback", "customer_satisfaction_score", "net_promoter_score", "brand_awareness",
    "market_trend", "consumer_preference", "survey_responses", "focus_group_feedback",
    "competitor_analysis", "advertising_cost", "customer_lifetime_value",

    # Tech & Software Development Fields
    "feature_flag", "bug_id", "bug_status", "release_version", "sprint_number", "repository_name",
    "commit_id", "branch_name", "build_number", "ci_cd_status", "deployment_status",
    "integration_test_results", "unit_test_coverage", "code_review_status", "pull_request_id",
    "api_version", "webhook_event", "cron_job_status", "job_execution_time", "server_load",
    "api_call_count", "database_query_time", "cache_hit_ratio", "error_stack_trace",

    # Additional General Fields
    "reference_number", "support_ticket_id", "customer_id", "user_experience_score", "support_case_status",
    "workflow_step", "escalation_status", "notification_status", "usage_statistics", "service_quality_score",
    "event_category", "configuration_settings", "feature_usage", "time_spent", "page_load_time",
    "interaction_count", "process_status", "feedback_status", "survey_completion_status", "resource_allocation"
]

hipaa_variations = set()
for field in hipaa_field_names:
    hipaa_variations.update(generate_variations(field))

non_hipaa_variations = set()
for field in non_hipaa_field_names:
    non_hipaa_variations.update(generate_variations(field))

df_hipaa = pd.DataFrame({
    "column_name": list(hipaa_variations),
    "is_hipaa_sensitive": [1] * len(hipaa_variations)
})

df_non_hipaa = pd.DataFrame({
    "column_name": list(non_hipaa_variations),
    "is_hipaa_sensitive": [0] * len(non_hipaa_variations)
})

df_combined = pd.concat([df_hipaa, df_non_hipaa], ignore_index=True)
df_combined = df_combined.sample(frac=1).reset_index(drop=True)
df_combined.insert(0, "id", range(1, len(df_combined) + 1))

csv_filename_variations = "dataset.csv"
df_combined.to_csv(csv_filename_variations, index=False)

print(f"CSV file '{csv_filename_variations}' has been created successfully!")