SparkCognition, identified on Google as “The Cognitive Security Analytics Company,” is an Austin-based start-up that is applying IBM Watson technology to the solution of cyber-security problems. SparkCognition Senior VP Usman Shuja spoke today with Etopia News to discuss how it does this.
He began by explaining that it keeps its massive security corpus/repository/data base current by constantly searching for new information about cyber-threats and how to remediate them. Humans play a significant role in finding new threads about threats and other security-related data for Watson to ingest.
Shuja said that his company’s SparkSecure system, powered by IBM Watson, is ideally suited to deal with the kinds of cyber-attacks on commercial and governmental systems that have been so much in the news of late.
Starting with proprietary algorithms for detecting anomalous, malicious intrusions into the computer systems it is protecting, SparkSecure also uses other techniques, including “machine learning” to constantly upgrade its own capabilities. This currently takes the form of “supervised learning” in which humans provide feedback to Watson, so it can learn from its mistakes, as well as strengthen and reinforce the correct actions it has been taking in certain circumstances.
Watson attaches a “confidence level” to its recommendations, which allows its human users to assess the relative usefulness of various suggestions.
The combination of SparkSecure and Watson enables the use of “predictive analytics,” wherein Watson, using the massive security repository it is operating with, can identify threats in large volumes of data, and deliver signatures of those threats to the IT staff responsible for system security. SparkSecure has successfully discovered several new threats.
According to Shuja, an average of fifteen percent of a company’s computer traffic is malicious. SparkSecure, he says, can identify that traffic and protect against it, thereby reducing the strain on system resources that these threats create.
Asked about the possible use of Watson to parse and analyze legal material, he said that this was a use case similar to what Watson is already doing in the medical field, where it diagnoses disease on the basis of the data presented to it and its understanding of a massive corpus of medical information, greater than what any one person could assimilate and recall. He said SparkCognition had no plans for such a general legal system, but that it did have a goal of helping a client develop a system for Watson-mediated compliance assistance on its road map.
He further explained that the Watson Ecosystem was a program to involve developers in the discovery of new use cases for Watson. He said that SparkCognition was a Premium Partner in this ecosystem, and that it was the only Watson-based company working on cognitive solutions against cyber-attacks. “Nobody else,” he said, “is using Watson for cyber-security.”
He pointed out that companies using SparkSecure have not yet reported any security breaches.
He also mentioned that using SparkSecure can reduce 10 hours spent doing remediation to 8 hours.
He emphasized that the SparkSecure model means that no client data is shared, but that successful remediation strategies used in one instance are added to the security repository for re-use in new situations.
In Machines of Loving Grace: The Quest for Common Ground Between Humans and Robots, author John Markoff distinguishes between AI and IA, Artificial Intelligence that does work by itself, and Intelligence Augmentation technologies, that assist humans in doing their work better. Shuja made it clear that SparkSecure “is an augmentation use case.” He said the system allows for an “auto-remediation” mode, but that most of the company’s clients use it to assist them in their work, not to replace staff.
He said it was inevitable that malicious hackers would develop innovative new attacks, but that the company was constantly on the alert for them and was prepared to meet any such new challenges.
Asked about a potential IPO, the senior VP said that, based on the traction they were already getting, they ought to be able “to build a large, successful company.” He thought they could do this “in a couple of years or less.” He said it was possible that a larger company could acquire SparkCognition. “Time will tell,” he concluded.