Fighting Fraud in Cryptocurrency using Machine Learning

Coinbase is on the front-lines of discovering advanced cryptocurrency and payment fraud techniques. Hear about how they use machine learning to help them fight the war.

Engineers are constantly required to build sytems to protect against various kinds of risks. For instance, downtime or system failure are certain types of risks. Protecting against a user's identify being stolen is yet another kind of risk. Engineers are constantly being required to to think ahead and to imagine not only the intended use-cases of a system, but also its mis-use or abuse cases. Soups Ranjan is an engineer who is no stranger to building complicated systems to mitigate risk. For his PhD thesis back in 2006, he designed a resilient system that could self-heal when responding to DDoS attacks.

Meet Soups Ranjan

Soups Ranjan is the Director of Data Science at Coinbase, the largest bitcoin exchange in the world. He manages the Risk & Data team that is chartered with preventing avoidable losses to the company due to payment-fraud or account-takeovers, which he will discuss in greater detail at DataEngConf. 

These days, Soups is taking on both known and unknown threats against cryptocurrencies in his role at Coinbase, the digital currency exchange that stores $1B in virtual currencies on behalf of its users. This huge pile of cryptocurrency attracts a plethora of bad actors and fraudsters who are constantly trying to hack the horde.

Compared to fiat currency, digital currency offers a highly unique set of security challenges because it is instantly transferable and transfers cannot be reversed. This makes the challenges of payment fraud and user security problems at Coinbase harder than any other company in the payments space. Coinbase is like the canary in the coal mine since bad actors try their latest techniques against digital currency exchanges first.

In his talk at DataEngConf SF, Soups will recount stories of some of the wildest scams he's seen and examples of attacks that he's faced in his role including phone porting attacks, previously unknown two-factor-authentication attacks, hackers attempting to beat online ID verification and more.

Most importantly, Soups will describe the multifaceted system Coinbase has built to detect potential fraud and to shut out bad actors using a hybrid of machine and human interactions.

How does Coinbase keep ahead of the hackers?

Soups helped the company build a fraud-prevention system that utilizes humans and machines in different roles to interoperate together. Since using only machine or only human analysts to stamp payment fraud is not optimal, Coinbase discovered they needed both working in cohesion and supporting each other in a fast feedback loop. Humans label inputs while the machine learns. In this way, the company has built an active learning system that can effectively utilize machine learning as one piece of the puzzle to prioritize risky users to watch more closely.

Want to hear more talks like this? Join us at DataEngConf SF '18 and be part of the biggest technical conference around data science and engineering.