<!-- # Privacy and Auditability by Design: Re-Visiting Open Finance --> # Privacy Preserving AML- Transaction Network Graphical Analysis **By [Silence Laboratories](https://www.silencelaboratories.com/)** <!----> <!-- # --> <!--## Content 1. Abstract 2. Open Finance: Background 3. Open Finance: Challenges 4. Role of Privacy, Consent and Trust 5. PETs: MPC 6. Design Adaptation in Account Aggregator: Silent Compute 7. Technical details for MPC in above application 8. Achieving Regulatory Compliance 9. Value Propositions: Before Vs Now 10. Further Readings--> :::info :bulb: The AML Transaction Network demo illustrates how multiple banks can collaborate on their respective transactional data to detect money-laundering and other fraudulent networks, without exposing any underlying customer data. This is enabled by using cryptographic technologies powered by Multi-Party Computation (MPC). ::: This proposal outlines how a consortium of five banks can collaboratively detect financial fraud using Privacy-Enhancing Technologies (PETs), specifically Secure Multi-Party Computation (MPC). Traditional fraud detection is constrained by siloed data and regulatory limitations, leaving gaps that criminals exploit. By leveraging MPC, banks can run joint fraud analytics on encrypted data — pooling intelligence without sharing raw information. The network enables encrypted querying, collaborative scoring, and privacy-preserving alerts, allowing early detection of complex fraud schemes like mule networks, identity duplication, and transaction layering. The proposal includes an end-to-end detection flow. :::warning 🧠 **The Problem** 1. Fraud schemes increasingly span multiple institutions (e.g., mule accounts, cross-bank layering). 2. Traditional systems offer limited visibility and no safe way to query peer banks. 3. Sharing raw customer data is restricted by confidentiality laws and data protection regulations (e.g., GDPR). :::  :::success 🔐 **Role of Privacy-Enhancing Technologies (PETs)** 1. PETs allow banks to collaborate securely without exposing sensitive data. 2. MPC enables joint computations on encrypted data, ensuring no party sees others' inputs. ::: Key capabilities of a network based on PETs like MPC are encrypted query & response (e.g., Bank A queries others without revealing intent), joint scoring of shared risks and data non-disclosure throughout the workflow. The figure below, enumerates the steps involved for one such design to identify AML patterns  :::info **A. Demo video:** https://youtu.be/-_lv9bL0mbc **B. Interactive demo:** https://preview--shielded-transaction-insights.lovable.app/ ::: Following sections describe the invidula steps for privacy preserving AML- transaction network graphical analysis. ## Step 1: Connecting transaction database All the participating banks integrate the Cryptographic [Computing Virtual Machine (CCVM)](https://md.silencelaboratories.com/wt8AXb7MSgu2_rpknUvHog?both) endpoint at their end, which links to their transaction database but never moves any raw data outside the respective bank’s environment. The inputs can be a simple list of transactions, with information about the transacting parties, amounts, transaction dates etc. The requesting bank, regulator, or the intelligence unit also selects visualisation of a pre-defined AML typology across the network of transactions, or defines their own custom logic.  ## Step 2: Connecting the transaction database The query is compiled into opcodes that the [Cryptographic Computing Virtual Machine (CCVM)](https://md.silencelaboratories.com/wt8AXb7MSgu2_rpknUvHog?both) can execute across the network of banks. The CCVM runs these instructions through a series of communications between bank endpoints, ensuring that no participant ever learns another bank’s underlying data.  ## Step 3: Network graphs Upon completion, the system visualises the results as an interactive network graph. Nodes denote anonymized accounts, edges denote transaction flows, and the red color coding highlights clusters that breach the configured risk thresholds. Analysts can hover or click on any element to see aggregated statistics - total volume, transaction counts, risk scores - while preserving privacy. This bird’s-eye view reveals hidden money-laundering rings that span multiple institutions, insights that would be impossible for any single bank to visualise alone.  ### Sample Patterns #### A. Smurfing/Structuring Transactions :::danger Abdel Elgendy structured over $250,000 in deposits to evade reporting requirements ::: *Elgendy purposefully engaged in a pattern of structuring activity to deposit over $250,000 in cash at financial institutions in ways to avoid reporting requirements for transactions in excess of $10,000*  #### B. Circular Transactions :::danger Greg Lindburg guilty of a $2B money laundering scheme ::: *Lindberg created a complex web of insurance companies, investment businesses, and other business entities and exploited them to engage in millions of dollars of circular transactions.*  ## Step 4: Investigation and reporting Finally, the banks can drill down into any suspicious cluster to review anonymised summaries and decide on compliance actions - filing Suspicious Activity Reports (SARs), freezing accounts, or escalating for deeper investigation.  :::success ### Key Benefits ● **Privacy by Design:** There’s no movement of raw data outside a bank’s environment, ensuring full confidentiality of customer data throughout the process. ● **Superior Detection of Cross-Institution Networks:** Identify complex money-laundering rings that span multiple banks, patterns invisible to any single institution, via a unified, privacy-preserving analytics engine. By fostering trust and catalysing network effects, privacy technology enhances fraud detection, becoming more accurate as additional banks participate. ● **Regulatory Compliance & Auditability:** Audit trails and compliance reports with detailed computation logs can help with regulatory compliance, satisfying KYC/AML requirement. ● **Operational Efficiency & Cost Savings:** The CCVM orchestrates complex cryptographic workflows across a network of participants, slashing manual data-sharing overhead and delivering insights far faster than traditional consortium models and bilateral agreements. ::: ---- ### Silence Laboratories Silence Laboratories, a provider of full stack Cryptographic Computing Power (VM branded as Silent Compute), is a privacy tech company building infrastructure and libraries for businesses to collaborate and compute without ever needing to share raw data. Using Silent Compute enterprises can connect their data pipelines (either enterprise data or retail consumer data) and use marketplace to use hosted functions/build new programs to run the computing VMs and expose business inferences. You can imagine us as Distributed DataBricks or Distributed Data Lakes. **Team:** Founded by CS PhDs with academic affiliations at MIT, NUS, UIUC, SUTD, and IITs with expertise in applied cryptography, compiler design, and algorithms. Technical dev center in Eastern Europe, R&D at Boston, and a global business team, HQ in Singapore.