Decoding “ OceanBrain” — The Artificial Intelligence Securing OceanEx

TL;DR — OceanEx embraces cutting-edge artificial intelligence technology in order to secure user’s digital asset in their crypto exchange. The AI component of OceanEx called OceanBrain builds deep learning models from the temporal market data in OceanEx and uses them to detect outliers. Deep learning is currently among the most powerful machine learning paradigms and is being used extensively in our daily lives with engine and recommender systems, as well as cutting-edge research including autonomous driving.

Cryptocurrency Exchanges

Cryptocurrency exchanges provide user-friendly trading platforms while providing market liquidity. There are two types of exchanges: centralized and decentralized. Centralized exchanges are much like stock brokerages, which have custody of customers’ funds thereby achieving highly efficient trading relative to the decentralized exchanges. On the other hand, decentralized exchanges provide an exchange between two users via the blockchain by utilizing smart contracts. Since decentralized exchanges do not hold custody of a customer’s fund, it is believed by some that they tend to be safer than the centralized exchanges. This thought remained until a recent incident involving the decentralized exchange Bancor, which was hacked due to vulnerabilities in their smart contracts.

OceanEx is a centralized exchange that enables high-frequency and algorithmic trading taking full advantage of liquidity in off-chain cryptocurrency trades. Through compliance with strict financial regulatory requirements, OceanEx protects platform-users and their assets while also offering FIAT exchanges in order to meet the asset-management needs of customers.

Exchange Security

Exchanges have learned many lessons from these hacks. We maintain a majority of customer assets in cold storage, and the computer has the private key which will never be connected to the Internet thereby making it impossible to hack the cold storage. We implement a stringent and fine-grained access control policy of platform roles so that employees do not have access to extraneous confidential information. We also deploy a two-factor authentication process to block suspicious logins and protect our customer accounts.

Because these exchanges are in essence complicated software/web applications there can always be sparse undetected bugs and vulnerabilities, possibly due to code issues or misconfigured infrastructure. Aside from these relatively newer blockchain exchange companies, massive web breaches are happening to large companies that spend a tremendous amount of resources on security including Equifax and Yahoo, as well as the leak of hundreds of million voter records from the US Government.

Indeed vulnerabilities in Internet software are not unique to the crypto industry, but the anonymous nature of blockchain accounts makes things more challenging compared to other Internet software used in the traditional financial sector. Addresses and private keys can be generated by any computer, whereas money transferring across banks can easily be traced thanks to KYC and AML regulatory requirements of banks. The consequence? If the hackers manage to sign the transaction transferring money from an exchange into their accounts, they can safely sit on top of a big score while watching angry customers accusing the “inept” and “irresponsible” exchange (instead of the thieves worrying about how to get rid of the feds and making nefarious plans for money laundering).

OceanBrain - AI for Exchange Security

To tackle this challenge, OceanEx is empowered by OceanBrain: an artificial intelligence system that uses advanced deep learning algorithms to learn from a significant amount of market data stream, and adaptively discover abnormal patterns in the market. With OceanBrain all of the platform information such as trading transactions, user activities (e.g., login, withdraw and deposit), API activities, market movements, and infrastructure status all comprise a broad spectrum of statistical measurements. OceanBrain uses a proprietary convolutional neural network to learn from the time series of these measurements, discover sophisticated temporal patterns over these time series, and use the learned patterns to differentiate normal system and abnormal system (i.e. detect possible hacks). When the trends indicate that the cause of abnormal detection is attributed to a set of user accounts these will be suspended immediately (since they may be compromised). However, in certain scenarios, the patterns can become very complicated and hard to attribute to individual users, so the entire exchange may need to be temporarily halted unless the threats are cleared by our security team.


The AI-powered Next Generation Intelligent Digital Asset Exchange Ecosystem