Operations
12/12/2016
11:55 AM
Rutrell Yasin
Rutrell Yasin
Slideshows
Connect Directly
Twitter
RSS
E-Mail

5 Things Security Pros Need To Know About Machine Learning

Experts share best practices for data integrity, pattern recognition and computing power to help enterprises get the most out of machine learning-based technology for cybersecurity.
2 of 6

#1 Garbage In, Garbage Out

There is an old saying in machine learning called 'garbage in, garbage out,' says Cylance's Matt Wolff. Situations where machine learning would not be effective are places where not enough data is generated to provide insight into what is going on in the IT network, he says. 

'As long as data is present and useful, then machine learning can also be useful.  If the data isn't informative, then machine learning isn't going to work,' Wolff says.

'I am bullish on the machine learning opportunity, but nothing as far as I know has been vetted out to demonstrate that this technology is going to definitively be better than any other techniques we are using to try to detect attacks,' says Pete Lindstrom, research vice president for security strategies with IDC.

Inputs and Outputs 

Presumably, organizations want to use machine learning to quickly react to attacks. Yet, a whole plethora of diverse types of information feed into the security ecosystem - for example, network packet activity, system calls on endpoints, and user behavior data at a meta level on the network. So, first people need to figure out what information is feeding the system, Lindstrom says. 

'You still must understand the inputs and outputs.  What is the nature and type of data feeding the system? What process and techniques do they use to identify the algorithms?  What is the action they might take?' 


The challenge is that he nature of these techniques is so dynamic, cybersecurity analysts can't take for granted that their output is going to be the same as someone else's output, explains Lindstrom: 'You can't rely on efficacies without doing your own testing on the solution. If it is looking for anomalous activity on the network, the only way to determine what is anomalous is to learn your network. It can't take someone else's network and apply it. Otherwise we are almost back to signature-based defense.'

Added complexity 

If security professionals are not careful they can wind up increasing complexity with less understanding of the processes and the output is going to be the same as security tools are delivering today. 'It is not horrible, by the way, because the solutions we have today block a lot of attacks,' Lindstrom says. 

Image Source: Rawpixel.com via Shutterstock

#1 Garbage In, Garbage Out

There is an old saying in machine learning called garbage in, garbage out, says Cylances Matt Wolff. Situations where machine learning would not be effective are places where not enough data is generated to provide insight into what is going on in the IT network, he says.

As long as data is present and useful, then machine learning can also be useful. If the data isnt informative, then machine learning isnt going to work, Wolff says.

I am bullish on the machine learning opportunity, but nothing as far as I know has been vetted out to demonstrate that this technology is going to definitively be better than any other techniques we are using to try to detect attacks, says Pete Lindstrom, research vice president for security strategies with IDC.

Inputs and Outputs
Presumably, organizations want to use machine learning to quickly react to attacks. Yet, a whole plethora of diverse types of information feed into the security ecosystem - for example, network packet activity, system calls on endpoints, and user behavior data at a meta level on the network. So, first people need to figure out what information is feeding the system, Lindstrom says.

You still must understand the inputs and outputs. What is the nature and type of data feeding the system? What process and techniques do they use to identify the algorithms? What is the action they might take?

The challenge is that he nature of these techniques is so dynamic, cybersecurity analysts cant take for granted that their output is going to be the same as someone elses output, explains Lindstrom: You cant rely on efficacies without doing your own testing on the solution. If it is looking for anomalous activity on the network, the only way to determine what is anomalous is to learn your network. It cant take someone elses network and apply it. Otherwise we are almost back to signature-based defense.

Added complexity
If security professionals are not careful they can wind up increasing complexity with less understanding of the processes and the output is going to be the same as security tools are delivering today. It is not horrible, by the way, because the solutions we have today block a lot of attacks, Lindstrom says.

Image Source: Rawpixel.com via Shutterstock

2 of 6
Comment  | 
Print  | 
Comments
Newest First  |  Oldest First  |  Threaded View
JonKim
50%
50%
JonKim,
User Rank: Author
12/15/2016 | 3:02:27 PM
Insightful
Insightful, thank you for sharing.
gopinathmohan861
50%
50%
gopinathmohan861,
User Rank: Apprentice
12/14/2016 | 10:11:16 AM
Machine Learning - Useful points
First of all, a big thanks for the article. The informations (5 security pros) mentioned in this article very useful. As AI and ML is going to rule future world, we need to consider these security pros.
Register for Dark Reading Newsletters
White Papers
Video
Cartoon
Current Issue
5 Security Technologies to Watch in 2017
Emerging tools and services promise to make a difference this year. Are they on your company's list?
Flash Poll
New Best Practices for Secure App Development
New Best Practices for Secure App Development
The transition from DevOps to SecDevOps is combining with the move toward cloud computing to create new challenges - and new opportunities - for the information security team. Download this report, to learn about the new best practices for secure application development.
Slideshows
Twitter Feed
Dark Reading - Bug Report
Bug Report
Enterprise Vulnerabilities
From DHS/US-CERT's National Vulnerability Database
CVE-2013-7445
Published: 2015-10-15
The Direct Rendering Manager (DRM) subsystem in the Linux kernel through 4.x mishandles requests for Graphics Execution Manager (GEM) objects, which allows context-dependent attackers to cause a denial of service (memory consumption) via an application that processes graphics data, as demonstrated b...

CVE-2015-4948
Published: 2015-10-15
netstat in IBM AIX 5.3, 6.1, and 7.1 and VIOS 2.2.x, when a fibre channel adapter is used, allows local users to gain privileges via unspecified vectors.

CVE-2015-5660
Published: 2015-10-15
Cross-site request forgery (CSRF) vulnerability in eXtplorer before 2.1.8 allows remote attackers to hijack the authentication of arbitrary users for requests that execute PHP code.

CVE-2015-6003
Published: 2015-10-15
Directory traversal vulnerability in QNAP QTS before 4.1.4 build 0910 and 4.2.x before 4.2.0 RC2 build 0910, when AFP is enabled, allows remote attackers to read or write to arbitrary files by leveraging access to an OS X (1) user or (2) guest account.

CVE-2015-6333
Published: 2015-10-15
Cisco Application Policy Infrastructure Controller (APIC) 1.1j allows local users to gain privileges via vectors involving addition of an SSH key, aka Bug ID CSCuw46076.

Dark Reading Radio
Archived Dark Reading Radio
In past years, security researchers have discovered ways to hack cars, medical devices, automated teller machines, and many other targets. Dark Reading Executive Editor Kelly Jackson Higgins hosts researcher Samy Kamkar and Levi Gundert, vice president of threat intelligence at Recorded Future, to discuss some of 2016's most unusual and creative hacks by white hats, and what these new vulnerabilities might mean for the coming year.