IBM Surveillance Insight for Financial Services:
V1.0 (Released: 12/22/2015)
IBM Surveillance Insight for Financial Services delivers a single cognitive solution to meet the regulatory needs of the world’s largest brokerage firms. By accumulating and analyzing vast quantities of data, both structured and unstructured, the system monitors anomalous behavior within a user group and identifies rogue trades for review by a compliance officer. Using a single comprehensive and holistic approach the system monitors fundamental precursors for each violation type. Utilizing a wider spectrum of inputs than traditional systems this cognitive approach offers both improved accuracy and greater efficiency the compliance officer.
IBM Surveillance Insight for Financial Services is a cognitive and holistic solution for monitoring all trading-related activities. It improves upon current surveillance process results and delivers alerts with greater efficiency and accuracy.
Surveillance Insight for Financial Services is based on a reasoning engine that goes beyond a traditional rules-based alert system. This holistic solution integrates information from multiple data sources, such as trading data and electronic communication, resulting in learning models that are continuously updated. It learns anomalies from data, by comparing traders’ behavior with self, population, and affinity groups. This cognitive reasoning engine also hierarchically correlates the causality of semantic events, to predict the potential upcoming risk. Information and alerts are delivered through effective and easy-to-use visualization tools.
IBM Surveillance Insight Advanced Analytics Include:
The surveillance capabilities help identify new patterns by linking various information with trader profiles, enabling you to "know your trader."
By deploying Surveillance Insight for Financial Services, you can improve the accuracy of alerts, reduce false positives and negatives, and improve the efficiency of investigations.
IBM Surveillance Insight Modules
1. Cognitive Reasoning Engine
At the heart of the Surveillance Insight platform is the "Cognitive Reasoning" engine that enables the system to fuse multiple independent data sources and reason through low-level alerts. By leveraging a Markovian Bayesian reasoning network the system is able to effectively scrutinize each new piece of information both within context to a base reference group and an individual's own historic behavior. Because each threat scenario is fundamentally unique this reasoning network can be optimized and fine-tuned for to include additional data sources or analytics as needed. By utilizing a cognitive system which can both "remember" and effectively "reason" through a known violation type. Compliance offers can achieve greater efficiency by reviewing low-level alerts in aggregation and empower better decisions with the use of more data in context.
2. Know Your Trades + Graph Querying
The graph modeling capabilities of the Surveillance Insight platform enable a Compliance Officer to explore and query relationships between flat relational datasets that were traditionally reserved for excel spreadsheets. By encoding information into the graphical model, we are able to achieve both an elegant and simplistic design while delivering copious amounts of valuable data that the user's fingertips. Take for example the Trader-Trader centric view where we can display a group of individuals connected by the number of common securities that they traded on any given day. Hovering over the edge of that connects the users we can see which securities are shared in common. Double-clicking on an individual or security we can query what is known as the "Ego-network" of that entity and display all the users who traded that security and what other securities were traded by those users. This powerful interface displays some of the fundamental benefits achievable through the use of a Graph database.
3. Live Monitoring & Replay
In the Live Monitoring and Replay functions we allow a Compliance officer to both monitor live transactional as they are processed into the system in real-time or query and replay and point in time in the systems history. Because this interface can be launched in a browser window a Compliance officer can easily launch multiple Live monitoring sessions and create preset filters to display only certain users, securities, or groups of users. Similarly the Replay function allows the Compliance officer to input any start-point and replay the transactions of that day while filtering by any group of users or securities.
4. Know Your Trader
The Know Your Trader module analyzes the electronic communication data of the underlying trader to build an emotional module for that trader. By applying the "Big 5 Personality" traits the system is able to identify key words and phrases in a users communication that can be organized into traits and emotions. These emotions break into five main factors: openness to experience, conscientiousness, extraversion, agreeableness, and neuroticism. By identifying these key emotions the Compliance officer can easily query and access communication logs that may have been identified as concerning or worrisome. These models can be further tuned and used as inputs into the Cognitive reasoning engine.
5. Trade Analysis
Fundamental to any Trade Surveillance platform is a market analysis tool that enables reviewed of structured alerts. In the Trade Analysis module the Compliance officer can review the trade anomalies that have been identified by violation type. Using an interactive interface the officer can see trade orders overlaid on the securities intraday activity as well as query all trades on a security within a given timeframe or window. By incorporating the underlying market data available to the client we offer advanced detection analytics and interfaces that help support the review of these structured trade anomalies.
6. Communication Analysis
The second main pillar of data used by the system is the "ECOMM" or electronic communication data. Whether this is email communication, Bloomberg chat communication, internal chat communication, or voice conversations that have been transcribed to text, the system is able to analyze a trader's communication while applying a domain specific lexicon to identify key points of interest within a user's communication. This module supports the Cognitive reasoning engine by generating the low-level alerts needed to contextually support the threat types as they are being reasoned.
Example Use Cases
Within the brokerage world there exists numerous known and unknown violation types. These prohibited activities can range from known SEC Violations such as "Pumping & Dumping" a security to unknown threats such as convoluted collusion scenarios where multiple trades work collectively to artificially manipulate a security in their favor. As or cognitive solution continues to iterate and be built out we plan to focus go beyond just known violation types but to mine user behavior to detect patterns which may prove to be illicit behavior.
In the video demo's below you fill find an example of how the Surveillance Insight for Financial Service's cognitive approach will fuse multiple data sources to tackle each of these violation types.
1. Pump & Dump Detection
Pump & Dump is a scheme that attempts to boost the price of a stock through recommendations based on false, misleading or greatly exaggerated statements. The perpetrators of this scheme, who already have an established position in the company's stock, sell their positions after the hype has led to a higher share price. This practice is illegal based on securities law and can lead to heavy fines.
The victims of this scheme will often lose a considerable amount of their investment as the stock often falls back down after the process is complete.
2. Collusion Detection
Collusion occurs when two traders agree to participate in a non-competitive arrangement that attempts to disrupt the market's equilibrium. By collaborating with each other, rival firms look to alter the price of a good to their advantage. The parties may collectively choose to restrict the supply of a good, and/or agree to increase its price in order to maximize profits. Groups may also collude by sharing private information, allowing them to benefit from insider knowledge.
Collusion involves people cooperating or working together when they should be competing. In the stock market, collusion can take many forms. Traders participating in accommodation trading, where goods are exchanged for non-competitive prices, are involved in collusion.
Colluding traders might share private information regarding upcoming takeovers, allowing them to benefit from insider trading. Price rigging also involves the collusion of sellers, who inflate the price of an asset to realize higher profits.