RADR(RADR) là gì?

ShadowPath is an open-source privacy protocol for Solana that enables unlinkable SOL transfers using zero-knowledge cryptography. The protocol implements a non-custodial mixer that breaks on-chain transaction graphs by decoupling deposit and withdrawal addresses. The system operates through three core components: an Anchor smart contract that manages deposits and verifies proofs, a Groth16 zk-SNARK circuit that generates privacy-preserving withdrawal proofs, and a multi-layer relay network that obscures transaction metadata. Users deposit SOL along with a cryptographic commitment into the contract's Merkle tree. Withdrawals require a zero-knowledge proof demonstrating knowledge of a valid commitment without revealing which one, allowing funds to be claimed at any address with no traceable connection to the original deposit. The protocol's relay network consists of 13 independent nodes organized into entry, middle, and exit layers. Deposits traverse 6-9 hops through this network with randomized routing and timing delays, creating over 3,600 possible path combinations. Each hop uses a unique wallet address, preventing blockchain analysis from correlating entry and exit transactions. The implementation includes a multi-split feature that divides large deposits into randomized smaller amounts, further defeating amount-based correlation attacks. Technical implementation uses the BN254 elliptic curve for proof generation, Poseidon hashing for Merkle tree construction (optimized for zero-knowledge circuits), and a 20-level tree supporting up to 1,048,576 deposits. The protocol prevents double-spending through nullifier tracking while maintaining complete privacy of the deposit-withdrawal relationship. All code is written in Rust using the Arkworks cryptography library and Anchor framework. The system is intended for research and educational purposes, demonstrating privacy-preserving techniques on Solana. It has not undergone formal security auditing and should not be used with significant funds on mainnet.