Machine Learning in cybersecurity: Cutting through the hype
It seems that everyone is talking about Machine Learning these days. In the cybersecurity market, ML is heralded as the newest weapon against increasingly sophisticated and targeted threats. Unfortunately, some companies are taking liberties with their definition of ML to capitalize on the hype. Here's a little insight into the fundamentals of ML to help you separate empty claims and buzzwords from real value creation when evaluating cybersecurity solutions.
Machine Learning: What it is and why it matters
ML is a field of study that gives computers the ability to learn without being explicitly programmed. It begins with a model, or algorithm, that transforms observations into decisions. For example, Netflix inputs the titles you watch and the feedback you provide into an ML model to make movie recommendations. The behaviors of mobile device software can be used in a similar way to predict a cybersecurity threat. In both cases, the machine is "trained" to learn from patterns and trends within very large data sets to make predictions.
"The secret ingredient is not the model, it’s the data."
Data is the secret ingredient
Open-source tools, such as Google's TensorFlow and Amazon Machine Learning (AML), make it easy to build an ML application in an afternoon. But making ML effective in the real world is hard. That's because the secret ingredient is not the model, it's the data. In other words, a unique and proprietary dataset delivers an unfair advantage. For example, only Uber has access to the pickup and drop off points of every rider it transports. Similarly, only LinkedIn has access to its social graph (which Microsoft valued at $26.2 billion when it acquired the company). These proprietary datasets create enormous value for their owners and barriers for competitors.
The emerging autonomous vehicle market provides another example of the competitive advantage gained from data. Tesla has more than 200,000 vehicles on the road logging over 7 million miles every day and Elon Musk has forecasted production of an additional 500,000 cars in 2018. Because all Tesla vehicles are equipped to collect and transmit sensor information, each mile driven captures valuable data that can be used to train and improve Tesla Autopilot algorithms. This gives Tesla a huge advantage over other automakers and tech companies working on autonomous vehicles. Don't you want the best trained algorithms behind your self-driving car when you take your hands off the wheel?
World's largest mobile security dataset
Lookout benefits from a similar advantage. Over the past 10 years, we've built an enormous customer base with more than 125 million registered users in over 150 countries. In fact, we continue to grow our user network by more than 50 thousand new mobile devices every day. This massive global device network generates billions of unique and proprietary data records that no competitor can match. With the world's largest global dataset for mobile, we're able to train our proprietary machine learning model to identify new and advanced threats with remarkable accuracy.
While ML and its benefits are real, few companies have access to the truly large-scale datasets necessary to bring its advantages to market. Evaluating the source of real world data that a company has access to, especially those which competitors can't access, should be your starting point when evaluating mobile security technology.
Interested in learning more about Machine Learning? Watch this myth-busting Machine Learning webinar from Lookout VP of Security Intelligence Mike Murray and co-founder Kevin Mahaffey.