Dark Reading is part of the Informa Tech Division of Informa PLC

This site is operated by a business or businesses owned by Informa PLC and all copyright resides with them.Informa PLC's registered office is 5 Howick Place, London SW1P 1WG. Registered in England and Wales. Number 8860726.

Vulnerabilities / Threats

11/11/2015
03:30 PM
Mike Paquette
Mike Paquette
Commentary
Connect Directly
Twitter
LinkedIn
RSS
E-Mail vvv
50%
50%

Machine Learning: Perception Problem? Maybe. Pipe Dream? No Way!

Guided by an organization's internal security experts,'algorithmic assistants' provide a powerful new way to find anomalies and patterns for detecting cyberthreat activity.

Machine learning has a perception problem. I recently met with a public company CEO who told me that "machine learning" has become an overused buzzword just like "big data" was a few years ago. Only it's even worse with machine learning because no one really understands what it means.

In the most common misperception, machine learning is thought to be a magic box of algorithms that you let loose on your data and they start producing nuggets of brilliant insight for you. If you apply this misperception to the use of machine learning for cybersecurity, you might think that after deploying machine learning, your security experts will be out of a job since algorithms will be doing all their important threat detection and prevention work.

[Read why Simon Crosby thinks Machine Learning Is Cybersecurity's Latest Pipe Dream.]

In Simon's commentary, he argues (three times, even) that experts are a better choice than ML/AI (Machine Learning/Artificial Intelligence) for cybersecurity. But why choose between experts and machine learning at all? A more enlightened understanding of machine learning in cybersecurity sees it as an arsenal of  "algorithmic assistants" to help the security expert automate the analysis of data by looking for helpful anomalies and patterns -- but under the direction of the security experts. 

Here's an example: A security expert doing malware research reads an article that contains an analysis of a version of the infamous Framework POS malware that exfiltrates data over the DNS protocol. Knowing what kind of security infrastructure is already in place, she thinks, "Hmm, if that exfiltration was done slowly enough on our network, I'm not sure we’d be able to detect it." Thinking a bit more, "Wow, I can really see how it could take some organizations months to detect a data breach that uses this method!"

She then configures her machine learning software to continually analyze DNS requests coming from all clients (POS and workstations) on their network, instructing the machine learning algorithms to create baselines of normal DNS request activity sent from each client, and to perform a population analysis across all clients in case some machines are already performing exfiltration when the analysis starts. The machine learning engine starts this analysis, and gives her an alert any time unusual behavior indicative of DNS "tunneling," is detected. 

In this way, our security expert has just put one "algorithmic assistant" to work for her. It never sleeps, eats, or takes vacation, and it does exactly what she told it to do! Tomorrow, she thinks, "I'll figure out a way to put another algorithmic assistant to work looking for unusual SSH sessions, another issue I've been losing sleep over."

Machine Learning Algorithmic Assistants Have Several Skills
Almost all algorithmic assistants that utilize unsupervised machine learning have several skill sets based on modern data science. They can baseline normal behavior by accurately modeling time series data (any series of data with a time stamp on it – usually log data from servers, devices, endpoints, and applications); they can identify data points that are anomalous or "outliers;" and they can score the level of anomalousness of these outliers. Generally, you'll hear this set of skills packaged up under the term "machine learning anomaly detection."

More recent developments in machine learning-based security analytics have additional capabilities; think of these as "senior algorithmic assistants" that can take the work of their subordinate assistants and perform advanced functions such as influencer analysis, correlation, causation, and even forecasting, to provide even more context for the security experts.

Perception Problem: Maybe. Pipe Dream: No!
Here's an interesting data point: In an April 2015 survey performed by Enterprise Management Associates, for the second year in a row security analytics (Advanced Security/Threat Analytics & Anomaly Detection) scored in the top ranking for perceived value when compared to total cost of ownership (TCO), beating out 15 other security technologies.

For forward-thinking security pros, this kind of security analytics, powered by machine learning, is no pipe dream – and it's so much more than just marketing spin. It's a practical way to use newer technology to automate the analysis of log data to better detect cyberthreat activity, under the direction and guidance of an organization's security experts.

Mike has more than 30 years of technology product development experience, including executive roles with several startups in the areas of consumer apps, mobile app ecosystems, and Security Information and Event Management (SIEM). Previously, he spent more than a decade in ... View Full Bio
 

Recommended Reading:

Comment  | 
Print  | 
More Insights
Comments
Newest First  |  Oldest First  |  Threaded View
COVID-19: Latest Security News & Commentary
Dark Reading Staff 9/25/2020
Hacking Yourself: Marie Moe and Pacemaker Security
Gary McGraw Ph.D., Co-founder Berryville Institute of Machine Learning,  9/21/2020
Startup Aims to Map and Track All the IT and Security Things
Kelly Jackson Higgins, Executive Editor at Dark Reading,  9/22/2020
Register for Dark Reading Newsletters
White Papers
Video
Cartoon
Current Issue
Special Report: Computing's New Normal
This special report examines how IT security organizations have adapted to the "new normal" of computing and what the long-term effects will be. Read it and get a unique set of perspectives on issues ranging from new threats & vulnerabilities as a result of remote working to how enterprise security strategy will be affected long term.
Flash Poll
How IT Security Organizations are Attacking the Cybersecurity Problem
How IT Security Organizations are Attacking the Cybersecurity Problem
The COVID-19 pandemic turned the world -- and enterprise computing -- on end. Here's a look at how cybersecurity teams are retrenching their defense strategies, rebuilding their teams, and selecting new technologies to stop the oncoming rise of online attacks.
Twitter Feed
Dark Reading - Bug Report
Bug Report
Enterprise Vulnerabilities
From DHS/US-CERT's National Vulnerability Database
CVE-2020-15208
PUBLISHED: 2020-09-25
In tensorflow-lite before versions 1.15.4, 2.0.3, 2.1.2, 2.2.1 and 2.3.1, when determining the common dimension size of two tensors, TFLite uses a `DCHECK` which is no-op outside of debug compilation modes. Since the function always returns the dimension of the first tensor, malicious attackers can ...
CVE-2020-15209
PUBLISHED: 2020-09-25
In tensorflow-lite before versions 1.15.4, 2.0.3, 2.1.2, 2.2.1 and 2.3.1, a crafted TFLite model can force a node to have as input a tensor backed by a `nullptr` buffer. This can be achieved by changing a buffer index in the flatbuffer serialization to convert a read-only tensor to a read-write one....
CVE-2020-15210
PUBLISHED: 2020-09-25
In tensorflow-lite before versions 1.15.4, 2.0.3, 2.1.2, 2.2.1 and 2.3.1, if a TFLite saved model uses the same tensor as both input and output of an operator, then, depending on the operator, we can observe a segmentation fault or just memory corruption. We have patched the issue in d58c96946b and ...
CVE-2020-15211
PUBLISHED: 2020-09-25
In TensorFlow Lite before versions 1.15.4, 2.0.3, 2.1.2, 2.2.1 and 2.3.1, saved models in the flatbuffer format use a double indexing scheme: a model has a set of subgraphs, each subgraph has a set of operators and each operator has a set of input/output tensors. The flatbuffer format uses indices f...
CVE-2020-15212
PUBLISHED: 2020-09-25
In TensorFlow Lite before versions 2.2.1 and 2.3.1, models using segment sum can trigger writes outside of bounds of heap allocated buffers by inserting negative elements in the segment ids tensor. Users having access to `segment_ids_data` can alter `output_index` and then write to outside of `outpu...