Endorsing Security
After successfully launching the Endor Protocol and bringing the business world one step closer to advanced, AI-based prediction capabilities, we are leveraging our democratized data science platform to serve national security.
Data Science Made Accessible
Replacing the current Machine Learning paradigm, our proprietary technology pushes AI even further by applying Social Physics theories that are based on years of research carried out by MIT. Simply put, Social Physics uses a new social theory of human behavior that predicts future choices through behavioral commonalities. So, instead of building a data model for each predictive question, the technology uses numerous data sets to continuously identify millions of dynamic behavioral patterns that bind objects together. These patterns can then be used to find ‘lookalikes’ and answer predictive questions.
Through an automated engine that can answer hundreds of lookalike-type questions, the protocol enables the automation and democratization of Al and Data Science, allowing companies to go from asking a limited number of costly predictive questions, to having constant, effortless and affordable access to unlimited advanced answers.
Predict or Discover
Our technology supports two main types of query — predictive and discovery — each serving multiple use cases for various industries. To answer predictive queries, the platform is ‘fed’ data relating to people with known past behavior matching the desired use case — people who have taken a loan in the past 3 months, people who have purchased a product recently, and so on. Their behavioral patterns are then used to rank any given population, according to their tendency to show the same behavior in the future.
Prediction use cases are structured in the following manner:
To answer discovery queries, the platform is introduced to a known group of people that are taking part in a specific activity. Their behavioral patterns are then used to rank any given population, according to their tendency to be related to this activity, widening the circle of known entities in the group.
Discovery use cases are structured in the following manner:
Business Smarts in the Interests of Security
Following the successful penetration of accessible data science into the business world, we are now applying these tools for the benefit of national security. To prove its predictive capabilities in this area, our technology was used to analyze terrorist threats based on Twitter data. Like finding a needle in a haystack, the platform was faced with the challenge of identifying members with the same activity profile as the small group of known individuals.
We were given millions of hashed raw Twitter Tweets containing examples of 50 Twitter accounts of identified ISIS activists, and an additional 74 known ISIS accounts with identifiers extremely well hidden in the metadata.
It only took 24 minutes to identify the first 80 ‘lookalike’ ISIS accounts, more than half of which were amongst the 74 well-hidden accounts named by our client. The entire process did not compromise global privacy and data security regulations, as the entire dataset was fully encrypted, and no semantic information was shared during the process. The final output, delivered within two hours, included a list of the top 200 accounts most likely to belong to ISIS activists. The top 50 detected ISIS accounts included 35 of the known hidden accounts, the top 100 included 51, and the top 200 included 72 of all 74 hidden accounts. The short analysis time and very low false-positive rate demonstrated extremely high efficiency in carrying out security-related predictive queries.
Mutual Benefit
Though use cases vary, we have already seen the benefits of utilizing the protocol for both business and security-related predictions, and believe that the extension of our service range will create new mutual benefits that we can’t even predict at this stage… but our platform is on the case…
Onwards and upwards,
The Endor Team
