Deprecated: jetpack_lazy_images_blacklisted_classes is deprecated since version Jetpack 8.7.0! Use jetpack_lazy_images_blocked_classes instead. in /home/dbslmic1/public_html/wp-includes/functions.php on line 5088
As it becomes painfully clear that traditional cybersecurity solutions leave gaps, however small, that can be successfully exploited by adversaries, many users take heart when new and promising security solutions emerge. Technologies such as heuristics, deep packet inspection or behavioral analysis have brought hope of better protection in their time. Today, the latest trend in cybersecurity is artificial intelligence, and specifically machine learning (ML). The latter has been touted as the new remedy to security issues. However, a major challenge with ML is that due to its complexity, it’s difficult for security professionals to truly evaluate the use and effectiveness of ML technology in security products. As stated by Dan Ariely, the James B. Duke Professor of Psychology and Behavioral Economics at Duke University’s Fuqua School of Business, “Everyone talks about it, (but) nobody really knows how to do it. Everyone thinks everyone else is doing it, so everyone claims they are doing it.”
WHAT IS MACHINE LEARNING?
It’s been a few years since John Hennessy, past president of Stanford University, declared that “Machine Learning is the hot new thing.” In fact, it’s taken 60 years for ML technology to achieve that status. A similar concept, artificial intelligence (AI), seems to be following the same path and often, the terms AI and ML are used interchangeably . In reality, ML is a subset of AI, which covers a broad area of data analysis that enables algorithms to make decisions on their own by learning from data.
Data scientists have made huge progress since ML’s humble debut in the 1950s, when it took a room full of computers to teach a machine how to play checkers. Today, ML has permeated our everyday lives so deeply that we commonly use it without even knowing it: every time we receive movie recommendations or shopping suggestions, for example, or when a credit card company alerts us of a potential fraud.
A Quick Definition
Machine learning is a subset of the broader field of artificial intelligence (AI). ML teaches a machine how to answer a question or how to make a decision on its own. It contrasts with traditional programming, which requires giving a machine explicit instructions for it to answer specific questions. In fact, every imaginable case has to be programmed ahead of time in order to cover all possible situations.
For example, imagine you wanted to take a multiple choice test. You could memorize all the correct answers by heart, which would be the equivalent of traditional programming, or you could learn to understand the concepts behind the questions, and then use that knowledge to determine the correct answer. The latter method represents the fundamental principle of ML.
The important difference is that ML teaches a machine how to predict an answer. This offers many advantages, but the biggest is the ability for the machine to respond to situations that it has not specifically encountered before, replacing processes that would have required arduous and time-intensive human analysis.
A Vast Field
In a nutshell, ML learns by being fed multiple examples in the form of a dataset, and rules or algorithms to apply to that dataset. The more examples the machine sees, the better it can learn.
There are multiple types of ML and each works very differently. If we generalize the field, we can define three main categories of ML: supervised learning, unsupervised learning and reinforcement learning.
In supervised learning, the machine is trained using sample data that is labeled to tell the machine what the data represents. In other words, it knows what it’s looking at (the input) and it knows what answers are expected (the output). Based on that training, the machine should be able to analyze new data and predict the correct answer. Supervised learning has applications such as disease diagnostics, or speech recognition.
In unsupervised learning, the machine is trained using data that doesn’t have labels. That means that the machine does not know what the data represents nor what answers are expected. The machine will have to figure out on its own the patterns and structure of the unlabeled input and discover the expected output. The classification of movie genres in Netflix is an example of unsupervised learning
In reinforcement learning, the machine interacts with its environment to achieve a certain goal. It is similar to unsupervised learning, as the machine is trained using unlabeled data. However, in reinforcement learning, the machine receives feedback on the outcome. For example, a machine can use this model to learn how to play a game. If the machine receives positive feedback (it wins) or negative feedback (it loses) from the actions it takes, it will, over time, determine by itself the best strategy to win the game. Each victory will reinforce the validity of specific actions. Reinforcement learning applications are emerging in robotics for manufacturing.