The Promise of
Deep Learning for AML
The key insight of deep learning for AML is that deep neural networks (DNNs) can generalize from training data to identify patterns in transactions that are indicative of fraud. To be more specific, Generative Adverserial Networks (GANs) have been shown to be able to descriminate between normal money transfer patterns and anomalous money laundering situations by using historical patterns and graph-based models of money transfers. This makes it harder for money launderers to make small changes in how they transfer money to circumvent the existing static set of money laundering rules that are currently used to alert investigators of transactions potentially involved in money laundering.
Existing rule-based systems for identifying transactions that may involve money laundering consist of thousands of rules. They generate an alert when a rule matches for a transaction - suspecting the transaction to involve money laundering. These systems generate a huge numbers of false-positive alerts (alerts where the transaction did not involve money laundering) that take time and money to chase down.
Governance & Compliance
Hopsworks is built for Enterprises. Read the Product sheet for Hopsworks Enterprise to how it provides:
- TLS-Based Security for Data-in-Transit;
- Full Audit-trail support, Encryption for Data-at-Rest;
- Integration with Active Directory, LDAP, OAuth2;
- Project-based multi-tenancy, enabling data to be shared and processed in a cluster environment;
- Provenance support for Machine Learning Pipelines - enabling fully reproducible models;
- Conda environments & Pip Libraries in Air-gapped deployments;