Member 685014
Novice Member
- Joined
- May 13, 2014
- Messages
- 6
- Reaction score
- 2
- Points
- 28
When an independent organization testing security vendors claims malware has reached a whopping 400 million samples, traditional security mechanisms are deemed unfit. They are functionally obsolete in the face of the mass of new or never-before-seen malicious files.
With more than 390,000 new malicious programs emerging daily, traditional anti-malware engines that rely on signature-based detection are becoming not only inefficient, but also incapable of adapting and protecting users against new and advanced threats.
The alternative to traditional security sounds like something from SF movies – machine learning technologies capable of uncovering new or unknown threats accurately and extremely fast. However, it’s actually a clear-cut and reliable answer to today’s growing security risks.
How Machine Learning Works
When talking about machine learning we inadvertently end up mentioning artificial intelligence, but not in the way we have been indoctrinated by Hollywood. You’re not going to see Terminator robots visiting your office any time soon, especially if we’re talking about security-centric A.I.
Bitdefender has been developing machine learning algorithms since 2009 in response to the emerging rapid growth in malware samples. When applied to security, machine learning is all about learning how current known malware behaves and using that knowledge to identify the same behavior in files that have been seen or analyzed.
The idea behind machine learning is to design a system that can learn based on past experience and draw conclusions from premises known or assumed to be true. For instance, we all learn tons of stuff in school - from geography to math and physics. We’re all taught things we know to be universally true.
If at some point in our lives we were to be left stranded in a completely unfamiliar place, we would have to rely on our knowledge to cope with our surroundings. We would have to tap our mapping and geography knowledge to find a vantage point and establish our location based on landscape and relative position from various landmarks. The next obvious step would be to find which way is North. By tapping our high school physics, we know that if we have a paperclip and rub it to a piece of silk, it will instantly become magnetized. By placing it on a leaf floating on still water, the magnetized paperclip will point to magnetic north.
Machine learning works pretty much the same way -- by tapping a huge base of knowledge about malware behavior. It then uses that information to identify malicious behavior in new or unknown files.
Performance and Accuracy
One of the biggest challenges in designing fast and accurate security-centric machine learning technologies lies in balancing performance and accuracy. By leveraging cloud performance with machine learning capabilities and accuracy, Bitdefender has managed not only to enforce detection of new malware with over 99% accuracy, but also to extend that protection to users worldwide in a matter of seconds.
Imagine identifying a new malicious URL and, in as little as 3 seconds, sending that detection to the other side of the planet to protect someone who might inadvertently visit the same URL. Speed is essential when dealing with never-before-seen threats, while also keeping false-positives as close as possible to zero.
The infrastructure behind the Bitdefender cloud has been designed with a single mission in mind: performance and scalability. More than 500 million users worldwide rely on Bitdefender technologies to stay safe, so the cloud is constantly being fed malware telemetry that can reach a staggering 7 billion requests per day.
The immediate benefit of security-centric machine learning technologies is that they constantly learn from this vast pool of experience and use that knowledge to proactively combat never-before-seen threats.
Predicting the Future
A major benefit of machine learning in the security industry is that it is constantly trained to infer information from data streams to detect new threat patterns. This additive knowledge enables machine learning technologies not only to be up to speed with the newest threats, but also to proactively combat global online threats in a matter of seconds.
For instance, if you were constantly up to speed with emerging threats and completely understand how they behave, you get better at identifying new ones. The more information about malware fed into machine learning algorithms, the better they become at spotting advanced online threats and adding that information to their knowledge base. Technically, the learning process never ends, and this security-centric artificial intelligence will only become better at spotting threats.
With the evolution of malware and advanced online threats, machine learning offers the key toward a better, proactive approach to security. We’ve reached a turning point where an efficient security solution needs to rely on “smart” technology that can cope with the fast-paced and ever-evolving world of advanced online threats. Security-based AI is just that.
Find out more about Bitdefender’s newest AI technology
With more than 390,000 new malicious programs emerging daily, traditional anti-malware engines that rely on signature-based detection are becoming not only inefficient, but also incapable of adapting and protecting users against new and advanced threats.
The alternative to traditional security sounds like something from SF movies – machine learning technologies capable of uncovering new or unknown threats accurately and extremely fast. However, it’s actually a clear-cut and reliable answer to today’s growing security risks.
When talking about machine learning we inadvertently end up mentioning artificial intelligence, but not in the way we have been indoctrinated by Hollywood. You’re not going to see Terminator robots visiting your office any time soon, especially if we’re talking about security-centric A.I.
Bitdefender has been developing machine learning algorithms since 2009 in response to the emerging rapid growth in malware samples. When applied to security, machine learning is all about learning how current known malware behaves and using that knowledge to identify the same behavior in files that have been seen or analyzed.
The idea behind machine learning is to design a system that can learn based on past experience and draw conclusions from premises known or assumed to be true. For instance, we all learn tons of stuff in school - from geography to math and physics. We’re all taught things we know to be universally true.
If at some point in our lives we were to be left stranded in a completely unfamiliar place, we would have to rely on our knowledge to cope with our surroundings. We would have to tap our mapping and geography knowledge to find a vantage point and establish our location based on landscape and relative position from various landmarks. The next obvious step would be to find which way is North. By tapping our high school physics, we know that if we have a paperclip and rub it to a piece of silk, it will instantly become magnetized. By placing it on a leaf floating on still water, the magnetized paperclip will point to magnetic north.
Machine learning works pretty much the same way -- by tapping a huge base of knowledge about malware behavior. It then uses that information to identify malicious behavior in new or unknown files.
One of the biggest challenges in designing fast and accurate security-centric machine learning technologies lies in balancing performance and accuracy. By leveraging cloud performance with machine learning capabilities and accuracy, Bitdefender has managed not only to enforce detection of new malware with over 99% accuracy, but also to extend that protection to users worldwide in a matter of seconds.
Imagine identifying a new malicious URL and, in as little as 3 seconds, sending that detection to the other side of the planet to protect someone who might inadvertently visit the same URL. Speed is essential when dealing with never-before-seen threats, while also keeping false-positives as close as possible to zero.
The infrastructure behind the Bitdefender cloud has been designed with a single mission in mind: performance and scalability. More than 500 million users worldwide rely on Bitdefender technologies to stay safe, so the cloud is constantly being fed malware telemetry that can reach a staggering 7 billion requests per day.
The immediate benefit of security-centric machine learning technologies is that they constantly learn from this vast pool of experience and use that knowledge to proactively combat never-before-seen threats.
A major benefit of machine learning in the security industry is that it is constantly trained to infer information from data streams to detect new threat patterns. This additive knowledge enables machine learning technologies not only to be up to speed with the newest threats, but also to proactively combat global online threats in a matter of seconds.
For instance, if you were constantly up to speed with emerging threats and completely understand how they behave, you get better at identifying new ones. The more information about malware fed into machine learning algorithms, the better they become at spotting advanced online threats and adding that information to their knowledge base. Technically, the learning process never ends, and this security-centric artificial intelligence will only become better at spotting threats.
With the evolution of malware and advanced online threats, machine learning offers the key toward a better, proactive approach to security. We’ve reached a turning point where an efficient security solution needs to rely on “smart” technology that can cope with the fast-paced and ever-evolving world of advanced online threats. Security-based AI is just that.
Find out more about Bitdefender’s newest AI technology
Last edited by a moderator: