How does IoT security benefit from machine learning?

When computers and mobile devices run rich operating systems, there are a number of security solutions and encryption protocols that protect them from many threats when connected to the Internet. The Internet of Things is not the case.

There are currently billions of IoT devices in use, most of which have low-end processing power and storage capacity, and do not have the ability to scale security solutions. However, they are in a very harsh environment when connected to the Internet.

Basically, it's like not going to the armor to wear the armor.

That's why new IoT vulnerabilities continue to surface, and countless IoT devices are being attacked by hackers, botnets, and other evils every day. A malicious hacker can find thousands of vulnerable devices in the search engine Shodan in a matter of minutes. The lack of immune IoT devices often become the beachhead of more dangerous hackers on the network. The bottom line is that we have too many smart devices that are too slow to protect ourselves (and us) from cyber attacks.

But this vulnerability can be compensated for by machine learning and analysis, especially through developers and manufacturers.

How IoT security benefits from machine learning

IoT devices generate large amounts of data, and machine learning is used to analyze and read data to help improve efficiency and customer service, and reduce costs and energy consumption. The same device can be used in safety-related use cases, such as determining the behavior of a security device and general usage patterns to help detect and prevent anomalous activity and potentially harmful behavior.

Currently, some high-tech companies are drawing on this approach to provide solutions that enhance the security of the Internet of Things, especially in smart homes that do not define security standards and practices.

Use cloud computing to consolidate intelligence

“Currently, machine learning and behavioral analysis is one of the biggest trends in detecting everything,” said Alex Baran, chief security researcher at network security technology company Bitdefender. However, he stated that machine learning still has a long way to go and requires “a lot of research and innovation on development, implementation and testing algorithms.”

BitDefender's approach is to aggregate a cloud server database that relies on all terminals of the product; input data for analysis to determine patterns and live malicious behavior. “You collect all the traffic,” Baran said. “By cleaning and regulating it, learning it, seeing what devices the device communicates with, what other devices, how they usually connect to the Internet and devices, and choose Abnormal traffic."

Machine learning is very promising, but it is still in its infancy and there is still a long way to go.

BitDefender uses cloud intelligence and pattern recognition to control the Internet traffic of home networks and block connections to malicious URLs, malware downloads, and suspicious packets through local network analysis of endpoint security software and hardware. The use of cloud services enables companies to bring enterprise-class intelligence and consumer space protection.

Human assisted machine learning

“Machine learning is a key component of the development of artificial intelligence for IoT security,” said Uday Veeramachaneni, co-founder and CEO of PatternEx. "The problem is that the Internet of Things will be distributed on a large scale. If there is an attack, you must respond in real time."

Most systems that rely on machine learning and behavioral analysis collect information about the network and connected devices and then seek out everything that is not normal. The problem with this original method is that it produces too many false warnings and false positives.

The approach suggested by PatternEx is to develop a solution that includes machine learning and enhances its insights with human analysts to detect larger attacks. “The way to solve this problem in real time is to create a learning system that uses these outliers and requirements that people feed back,” explains Veeramachaneni. "Only humans can distinguish between malicious and benign. These feedbacks are returned to the system and a predictive model is created. The model can mimic human judgment - but this needs to be done under huge scale and real-time conditions."

This is particularly relevant to the IoT ecosystem, which involves a large number of devices, and real-time analysis of the massive amounts of data generated is beyond human capabilities.

PatternEx uses machine learning algorithms for anomaly detection and trains the model to be more accurate in real time. Training is done by any analyst who can discover new attacks. The system generates events with potential attacks. The analyst investigates the incident and determines if the system's assessment is correct. The system learns from experience and makes more accurate decisions the next time.

“This model helps improve the accuracy of threat detection and reduces the number of false positives over time,” Veeramachaneni said.

Take advantage of the limited capabilities of IoT devices

IoT devices are designed to perform a limited set of functions. So with machine learning and enough data, it's very easy to identify anomalous behavior. The idea was used by the startup technology company's Dojo-Labs lab to create a smart home Internet of Things security solution.

“When it comes to IoT devices, they are designed to do a very, very specific function,” said Yossi ATIas, co-founder and CEO of the company. "So, assuming we have a lot of users using the same camera or the same smart TV or the same smart alarm or smart lock, there is no real reason that a device will behave differently than other behaviors because they all run the same software, And this is not something the user can change."

The Dojo-Labs lab approach involves collecting metadata from different endpoints and defining the scope of behavior for each device type to be able to detect and block malicious behavior. Just as all solutions involve machine learning, the Dojo-Labs lab's model has been improved by collecting more and more customer data.

The solution includes a cobblestone device installed in the home network, allowing users to control devices and mobile applications that monitor network status and a cloud server, using proprietary statistical techniques and mathematical models on the cloud server, plus Machine learning algorithms to synthesize and analyze data.

There are some notes on machine learning.

Machine learning is promising, but it is still in its infancy and there is still a long way to go. Never think of it as a complete solution. "[Machine learning] will be almost everywhere," Veeramachaneni said. “In order to be secure in the enterprise or in the Internet of Things, you have to have powerful machines to organize data, analyze data, and find patterns in the data. But you also need human intuition to discover new attacks and train the system to stop these new ones. (and old) attacks."

Veeramachaneni called this combination "augmented intelligence", abbreviated AI, which is the convergence of human and machine power to defeat cyber threats. He said: "Machine learning and human beings cannot do it alone."

Our coin handling machines mainly include Coin Counting Machine, Coin Sorting Machine and Coin Sachet Machine

Coin counting machine models: HCS-3300, HCS-3500AH, CS-10S, CS-2000

Coin sorting machine models: CS-211S, CS-311S, CS-610S+PRO, CS-910S+, CS-600B, CS-211S, CS-311S, CS-610S+PRO and CS-910S+ have alloy sensor for counterfeit detection, CS-600B sorts coin by diameter

Coin sachet machine: pack the loosed coins into small plastic bag, compared with paper roll packing machine, it is more economic solution 

Coin Handling Machine

Coin Counter Counting Euro Coins,Coin Counter For Vending,Coin Counter Super Large Hopper,Super Heavy Duty High Speed Counter

Suzhou Ribao Technology Co. Ltd. ,