LOCARD: Agentic Framework for Blockchain Forensics
A new paradigm called Agentic Blockchain Forensics (ABF) models forensic investigation as a sequential decision-making process, moving beyond static inference pipelines. The LOCARD framework operationalizes this through a Tri-Core Cognitive Architecture that separates strategic planning, operational execution, and evaluative validation. It uses a Structured Belief State mechanism to enforce forensic rigor under explicit state constraints. Applied to cross-chain transaction tracing, LOCARD introduces the Thor25 benchmark dataset with over 15,000 labeled transactions. The framework is designed for dynamic and iterative investigations, addressing the complexity of blockchain forensics. This represents a shift towards more adaptive and rigorous forensic methods in the blockchain domain.
Key facts
- LOCARD is the first agentic framework for blockchain forensics.
- It introduces Agentic Blockchain Forensics (ABF) as a new paradigm.
- The framework uses a Tri-Core Cognitive Architecture.
- It incorporates a Structured Belief State mechanism.
- Applied to cross-chain transaction tracing.
- Introduces the Thor25 benchmark dataset with over 15,000 transactions.
- Models forensic investigation as sequential decision-making.
- Enforces forensic rigor under explicit state constraints.
Entities
—