Senzing uses entity resolution to find relationships in databases | Tech News
Cybercrime hurts businesses’ bottom lines. Damage costs from social engineering, ransomware attacks, and malware infections are expected to hit $6 trillion annually by 2021, according to CSO’s State of Cybercrime report. Even more alarming? Since January 2017, hackers have exposed some 1.9 billion records containing sensitive information, Privacy Rights Clearinghouse reports.
With the threat of catastrophic breaches looming larger every day, it’s no wonder that companies are turning to cutting-edge tools to preempt cybercriminals before they make headway against their defenses. One of those tools is from Senzing, a Venice Beach, California-based startup founded by former IBM chief scientist of context computing Jeff Jonas.
It’s a company with an interesting history. Jonas prototyped an entity resolution system in the 1990s — Non-Obvious Relationship Analysis (NORA) — that Las Vegas casinos used to find criminals and shady dealers, including the card-counting MIT students featured in the book Bringing Down the House and the movie adaptation 21. It was rebranded Entity Analytic Solutions in 2001 and sold to IBM in 2005, and in 2009, IBM relaunched the project as “G2.” A few years later, Senzing licensed G2 from IBM and recruited a few of the original researchers.
Senzing’s core product is an AI-powered Mac and Windows app for entity resolution (ER), the process that determines the relationships between people, objects, and metadata, or “entities,” in a given database. It creates what Jonas refers to as a “resume” of each entity that includes all the names it’s known by, all the places it has been, and the email addresses it has used, among other things.
“On your phone, you probably have duplicates,” Jonas told VentureBeat in a phone interview. “Imagine this problem for a bank … or a health care organization with hundreds of thousands or even millions of identities to manage.”
The app, which is free to use on databases with less than 10,000 records (up to $50,000 for databases with tens of millions of records), builds on years of research and millions of dollars in investments, said Jonas. He claims it’s the only AI-powered entity resolution tool that self-tunes and self-corrects in real time.
“We’ve worked hard to make it where it doesn’t need to have all the configurations exposed,” he said. “You don’t need to be an expert. It’ll run locally on a two-core laptop or in a cloud environment … and it scales all the way up to billions of records.”
To that end, Senzing ingests data in CSV (comma separated values) format from a variety of platforms, systems, and services. Supported sources include email clients (Microsoft Outlook and Gmail); customer relationship management, or CRM, systems (Salesforce and Microsoft Dynamics); direct marketing systems (Mailchimp and Zoho); web and ecommerce platforms (WordPress and Stripe); accounting and HR systems (Quickbooks and ADP); and spreadsheets and other files.
After the data is imported and mapped with annotations that describe the attribute of each field (i.e., “NAME_LAST” for a person’s last name), Senzing is primed and ready to perform entity resolution. It’s smart enough to handle phone numbers in multiple formats, missing address fields, partial names, and other forms of unstructured data.
“Addresses are messy. So are phone numbers — some have a country code and others don’t,” Jonas said. “That makes it difficult for companies to resolve identity.”
The process takes anywhere from a few minutes to hours, depending on the size of the database.
A bank might use entity resolution to find “non-obvious” connections in their data, Jonas explained, like helping determine whether one customer has five checking accounts or five customers with the same name and address have five separate accounts. And consumer goods vendors might use it to compare their manufacturing partners to a database of known bad actors, such as companies that use child labor or toxic chemicals.
But it’s also a tool in the fight against “bad guys,” as Jonas likes to call them. An early test with a banking organization seeded a database from World-Check, a company that tracks individuals who pose risks to the financial industry, with 1,387 fictional records containing data from 572 criminals. The system found 97 percent of them — and also discovered 127 relationships of which the bank was previously unaware.
To date, Senzing has worked with the Singaporean government to monitor vessels passing through the Strait of Malacca and Singapore Strait, and with The Pew Charitable Trusts to driver voter registration in the U.S. and ensure that state voter lists were accurate and up-to-date.
“A lot of companies that do [entity resolution] only make their technology available to the elite,” Jonas said. “We want to democratize ours.”