An advanced Web3 identity and security framework that bridges financial behavioral risk analysis with hardware-isolated biometrics and zero-knowledge cryptography. The system dynamically enforces step-up verification only when behavioral transaction risk exceeds safety thresholds, eliminating unnecessary user friction while ensuring privacy-preserving Proof of Personhood.
Evaluates 30-dimensional transaction feature vectors using a pre-trained Keras deep neural network.
Dynamically calculates risk scores to route transactions: low-risk transactions are pre-approved instantly, while high-risk transactions (T_risk >= 0.30) trigger an adaptive biometric challenge.
Extracts 512-dimensional embeddings via a FaceNet architecture and verifies identity using spatial Cosine Distance (D_cosine < 0.60).
Executes hardware-isolated HMAC-SHA256 salting inside a simulated Trusted Execution Environment (TEE) by binding silicon identifiers (device_uid) with raw biometric vectors.
Side-by-side JSON breakdown of zero-knowledge proofs and direct Solana Devnet blockhash verification links.
Automatically formats live witness inputs into Prover.toml and invokes the Noir Nargo CLI to compile and generate zero-knowledge proof artifacts (pi) via ACIR circuits without exposing raw biometric templates on-chain.
As blockchain ecosystems grow, the logic within the smart contracts presents a significant risk to institutional security. This project provides an AI-Driven Risk Engine designed to autonomously audit and secure decentralized financial transactions.
Integrated with a Solana-based Registry to record safety scores and audit metadata permanently, ensuring full transparency.
The engine moves beyond basic syntax analysis to identify high-level behavioral triggers and high-risk flags such as "FREEZE FUNDS" or "MINT_UNLIMITED".
By utilizing "Safety hashes", the framework reduces information asymmetry between contract creators and users, fostering trust in automated finacial systems.
As decentralized finance (DeFi) matures, "Bank Run" scenarios in stablecoin ecosystems present a systemic risk to capital stability. This project introduces a Behavioral Volatility Index (BVI) - a risk-adaptive tool designed to autonomously stabilize liquidity vaults during periods of extreme market trauma, such as the May 2022 Terra-Luna de-pegging event.
Conducted an exhaustive analysis of 1,076 high-frequency data points from the May 2022 crash to validate the "Solvency Gap" between fixed-fee protocols and BVI-driven adaptive models.
Developed a specialized Rust/Solana-based Smart Contract that dynamically adjusts redemption fees in real-time based on market sentiment and volatility scores, successfully extending the liquidity runway in simulated stress tests.
Engineered using the Anchor Framework (Rust Edition 2024), leveraging Solana's low latency for sub-second updates to the BVI state, ensuring that arbitrage attempts remain costly during high-volatility windows.
Integrated "Safety Hashes" and audit metadata to provide full transparency into the vault's risk-adjustment logic, reducing information asymmetry for institutional and retail users alike.
This project combines individual-level ACS data with high-frequency Optiver market microstructure data to construct a prototype behavioral-market matching framework. Individual characteristics are used to construct a model-implied behavioral-risk score, while market data are used to characterize contemporaneous market conditions through WAP, returns, and rolling realized volatility. These two components are then mapped into a contract-level risk architecture that generates prototype risk-adjustment and hedging scores.
Constructed model-implied behavioral-risk scores from approximately 1.49 million cleaned ACS observations, using age and personal-income characteristics as inputs.
Processed 130M+ Optiver observations using a disk-based DuckDB pipeline to derive volume-weighted average price (WAP), log returns, and rolling realized volatility.
Developed a contract architecture that maps heterogenous individual risk profiles to contemporaneous market volatility regimes.
Produced risk-adjustment and hedging metrics, mapping their interaction with core behavioral traits and market turbulence.
Built an out-of-core feature engineering and visualization workflow capable of processing multi-gigabyte datasets without memory constraints.
This project establishes a dynamic risk-mitigation framework that integrates real-time biological stress signals with financial leverage states. By automating a "Biological Circuit Breaker," the system prevents emotional panic trading and preserves protocol solvency during extreme market volatility.
Developed an automated engine that modulates portfolio leverage caps (e.g., dampening from 10x to 1x) based on real-time "Biological Calm" (RMSSD) levels.
Constructed a correlation matrix to map the relationship between biological stress and market volatility, proving that internal bio-signals act as independent, non-correlated risk indicators (Avg. Coupling: 0.03).
Applied regression analysis to quantify the "Cortisol-Beta" relationship, demonstrating that higher physiological resilience facilitates higher sustainable portfolio beta.
Utilized Seaborn and Matplotlib to visualize longitudinal trade sequence coupling and risk-dampening states, providing empirical proof of efficacy in capital preservation.