DBbun/500K_Crypto_v1.0

DBbun Crypto Synthetic Dataset

DBbun Crypto Synthetic is a large-scale, privacy-safe simulation of blockchain-style transactions.

The dataset was inspired by the paper: “Beyond Static Datasets: A Behavior-Driven Entity-Specific Simulation to Overcome Data Scarcity and Train Effective Crypto Anti-Money Laundering Models.”

No data were copied or extracted from that study. Instead, DBbun recreated its behavioral logic and structural principles to generate fully artificial yet statistically realistic records.

---

Summary of the Generated Dataset

Table	Approx. Rows	Description
transactions.csv	~500,000	One synthetic transaction per row
edges.csv	~10–15 million	Sender → receiver edges forming the transaction graph
accounts.csv	~100,000	Unique wallets and entities
labels_entities.csv	~100,000	Entity-level labels (licit / illicit + K-hop proximity)
labels_transactions.csv	~500,000	Transaction-level labels (benign / suspicious + heuristics)
stats.json	—	Summary counts and distributions

---

Files Included

File	Description	Rows (approx.)
transactions.csv	One row per synthetic transaction with timestamp, pattern, and fee.	500K
edges.csv	Directed sender → receiver edges forming the transaction graph.	10–15M
accounts.csv	Wallet-level entities with balances, lifetimes, and address counts.	100K
labels_entities.csv	Entity labels (licit / illicit) + graph proximity (K-hop).	100K
labels_transactions.csv	Transaction labels (benign / suspicious) + heuristic flags.	500K
stats.json	Summary statistics and distributions.	—

Each table is self-contained and can be joined using common keys such as tx_id (for transaction-level joins) and account_id (for entity-level joins).

---

Schema Overview

Transactions

Each row represents a synthetic transaction in a blockchain-style ledger.

Column	Description
tx_id	Unique transaction identifier
timestamp	UTC datetime
pattern	Transaction behavior type (regular, mixer, coinjoin, exchange_withdraw, fan_out, peel_chain, single_use)
num_inputs	Number of input addresses
num_outputs	Number of output addresses
total_in	Total input amount
total_out	Total output amount
fee	Transaction fee
tx_hash	SHA-256 hash (deterministic and reproducible)

---

Edges

Directed sender → receiver relationships that form the transaction graph.

Column	Description
tx_id	Transaction identifier
timestamp	UTC datetime
sender	Sending wallet or address
receiver	Receiving wallet or address
value	Amount transferred
pattern	Transaction behavior pattern

---

Accounts

Wallet-level entity table representing participants in the system.

Column	Description
account_id	Unique wallet identifier
entity_type	Category (e.g., exchange, mixer, mule, business, service, licit, nested)
first_seen	Earliest activity timestamp
last_seen	Most recent activity timestamp
current_balance	Remaining balance
n_addresses	Number of associated addresses

---

Labels — Entities

Entity-level labels describing risk, type, and graph proximity.

Column	Description
account_id	Wallet identifier
entity_type	Same as in accounts
entity_label	licit or illicit
k_hop_dist	Integer distance from an illicit entity
k_hop_label	Categorical proximity flag (within_2hop, far)

---

Labels — Transactions

Labels and heuristics associated with each transaction.

Column	Description
tx_id	Transaction identifier
pattern	Behavioral pattern
tx_label	benign or suspicious
is_fan_in	True if many inputs converge
is_fan_out	True if one input splits into many
is_roundish	True if rounded amounts observed
is_bursty_hour	True if occurs during a local burst hour
is_bursty_day	True if part of a burst day

---

Use Cases

• Graph Analytics — explore transaction flows, community detection, and centrality.
• Machine Learning — train models for illicit-activity or suspicious-transaction detection.
• Education — teach blockchain analytics and anti-money-laundering frameworks.
• Benchmarking — stress-test ETL, graph databases, and networ