What do your Facebook newsfeed, Amazon product recommendations and Siri voice recognition all have in common? They all use some form of machine learning to automate data correlation, identify patterns, and make changes based on newly learned data. But what exactly is machine learning?
Machine learning is when a program takes new data, learns from it and makes changes without being explicitly programed to do so. Machines perform data sequencing in an automated fashion, combing through sets of data searching for patterns and similarities. Once data patterns and predictive behaviors have been identified, rules must be implemented to take action on learned data. With machine learning, the machine is enabled to create or modify rules to further improve itself and accomplish its primary objectives.
The objectives can vary and have a variety of uses. Machine learning can be used to make routes faster and more efficient for various transportation methods; improve sales conversions through product recommendations or tailoring product content to direct purchasing decision; predict hospitalization based on physical behavior and patient data; improve patient diagnostics based on trends or areas of concern; or determining levels of risk for investments or insurance.
In the times before machine learning, a user must manually provide the program with new sets of rules for any new data in order for any action to occur, as well as decide on what the next step would be to act upon any new rules. With machine learning, the program creates algorithms, or a sequence of instructions, to execute in order to accomplish a desired end result as well as recommend and act upon any suggested next steps. Rather than having to manually parse through copious amounts of data, correlate patterns, create algorithms and execute across systems, users, utilizing automation and machine learning, have been able to improve efficiencies by being more granular and prescriptive, as well as alleviate workload.
There are a few conditions necessary for machine learning to occur:
1. Data must be consolidated into one place so the machine may be able to access all relevant data necessary to make a decision.
2. The appropriate structure must be in place to analyze the large amounts of data to identify similarities and patterns. Combing through significant amounts of data sets requires powerful computational processing.
3. Basic set of rules must be provided to the machine to serve as a basic guideline, as well as provide the desired outcome.
The caveat with machine learning is the significant level of trust the user must place in the hands of the machine. There is a certain level of fear and trust associated when it comes to machine learning – a trust that the machine will take the appropriate actions, for instance putting their lives in the hands of the Tesla Autopilot or the Google self-driving vehicles – and a a fear machines might evolve into a technological singularity. Watching artificial intelligence takeover movies such as iRobot, Ex Machina, WarGames and Tron will demonstrate this to an extend.
While this fear is understandable, it’s important to note that machine learning helps predict behaviors and recognize patterns in a way that humans cannot due to their limited compute capacity. Machines are able to recognize patterns and act much faster than a human. They learn from previous and new conclusions to further develop and improve it’s own algorithms. When massive data sets are present, it is impossible to create algorithms at scale. The only way to do this is to utilize machine to process data and learn from experience to improve upon itself. Additionally, although machine learning enables self-implied changes in order to accomplish a specific objective, the objective itself is always set by humans. With this, there must also be an implied trust in the goal set by the user.
Specific to cybersecurity, it is difficult to keep pace with the constant volume and increasing sophistication of threats and attacks. Cyberattacks have evolved to be heavily automated – 68% of respondents in a recent Ponemon study1 agree that automated tools have made it easier for attackers to execute a successful attack. Our unique approach to cybersecurity incorporates automation and machine learning allows us to get ahead of attackers. Machine learning can help accurately identify variations of known threats, identify patterns, predict the next steps of an attack and automatically create and implement protections across the organization in near real-time. With machine learning, successful cyberattacks can be prevented. To learn more about next-generation security platforms, visit https://www.paloaltonetworks.com/products/designing-for-prevention/security-platform.