FloCon 2019 has ended

Sign up or log in to bookmark your favorites and sync them to your phone or calendar.

General Session [clear filter]
Tuesday, January 8

9:00am EST

Cutting Through the Hype: How to Effectively Apply ML to Cybersecurity
Current cybersecurity challenges represent a machine-scale problem and large amounts of automation are required to solve it. Data scales will continue to grow, further compounding the challenge. Defenders need to use the internal network and host log data that is already at their disposal, cross-network and cross-host, to discover the presence of sophisticated adversaries. This talk will detail a machine-learning based approach for how to solve this difficult problem--automated internal network monitoring, with low false positive rates--to find sophisticated adversaries and their campaigns.

It will discuss the three fundamental requirements to achieve effective monitoring with a reasonable, practical amount of resources:
  1. Focus on the adversary campaign holistically: Using a campaign-oriented framework for monitoring also simplifies what needs to be monitored. You only monitor behaviors the adversary must perform, the ones they cannot avoid, to succeed in their mission. This reduces the noise, false positives and level of effort required by analysts.
  2. Automation, machine learning, and interpretability: It is not possible today to directly model the problem: the community does not have enough examples of “known bad” (identified, true APT campaigns) and networks are too complex and varied. To frame this as a machine learning problem, it needs to be broken down into multiple sub-problems of monitoring for individual surprising behaviors. E.g., is this an unusual number of pings? Is this an unusually large data movement?
  3. Adapt to ever-changing environments and adversaries: the training of models must be automatic and not require human intervention, meaning they must train on data in situ, must be retrained and updated frequently to stay relevant, support using a variety of raw data sources, and be easily updatable to account for the latest and greatest adversary tactics.

Any approach that lacks these necessary pieces will not scale to large networks or will lag behind evolving adversaries.

Attendees will Learn:
1) Knowledge of the ways ML can be effectively and ineffectively applied to the challenges of cybersecurity, so they are more educated on to evaluate different tools for their unique environments
2) A strategic understanding of how to frame the problem of advanced threat detection so that machine learning can be effectively applied
3) A more in-depth understanding of the core behaviors in the adversary campaign, and how that enables a reduction in false positives

avatar for Jason Kichen

Jason Kichen

Vice President of Advanced Security Concepts, ESentire
Jason Kichen serves as the Vice President of Advanced Security Concepts at eSentire; prior to its acquisition by eSentire, Mr. Kichen was the Director of Security Research & Operations at Versive Previously, Mr. Kichen spent nearly 15 years working in the U.S. intelligence community... Read More →

Tuesday January 8, 2019 9:00am - 9:30am EST
Grand Ballroom 300 Bourbon St, New Orleans, LA 70130

9:30am EST

Improved Hunt Seeding with Specific Anomaly Scoring
As the practice of hunting has spread through enterprise cyber security, interest in generalized anomaly detectors to seed hunts has also increased. The generally accepted premise seems to be that security events are rare and rare events are almost always anomalous. Therefore, if one seeds hunts with anomalous events, then the hunts are more likely to uncover activity of interest. However, implementing directed hunts in this manner requires the ability to define and detect anomalous behavior within complex systems. This is typically done probabilistically and there are several products available that employ machine learning approaches, such as neural networks, to define anomalous network activity. However, unsupervised learning approaches are inescapably plagued by high false alarm rates which, in turn, lead to analyst alert fatigue. To decrease false alarm rates, one may tune such products to only alert in cases of extremely anomalous events. But this still fails to address the heart of the problem which is that generalized anomaly detection for an entire network is probably not an optimal approach. Rather, defenders should develop a number of specific models. What is required is a scalable and repeatable framework for doing so. We present an open source approach for cyber security experts and data scientists/statistical engineers to collaboratively develop specific anomaly scoring models. Our approach utilizes a non-parametric kernel density estimator to evaluate the distributions of security logs. Once the desired distribution has been learned, analysts may use it to score records with a single-number, probabilistic measure of anomalousness. Logs can then be filtered based on this anomalousness score and rare events can be utilized to seed hunts. After sufficient validation, models may be transitioned to detectors which alert defenders when some criterion is met.

Attendees will Learn:
Attendees will be presented with a flexible, open source tool for non-parametrically modeling multivariate densities of network logs. Once constructed, such models can be utilized to score the anomalousness of log records and facilitate directed hunting. More subtly, attendees will gain insight into the potential benefits available through iteratively collaborating with statistical engineers/data scientists, such as the construction of highly customizable models for specific phenomena on specific networks.

avatar for Brenden Bishop

Brenden Bishop

Data Scientist, Columbus Collaboratory
Brenden Bishop is a data scientist at Columbus Collaboratory. He focuses on developing prototype solutions for network defenders and enterprise IT in a variety of problem areas, including network anomaly detection and active directory tidying. He is a graduate of The Ohio State University... Read More →

Tuesday January 8, 2019 9:30am - 10:00am EST
Grand Ballroom 300 Bourbon St, New Orleans, LA 70130