15
The Enterprise Immune System Machine Learning for Cyber Threat Defence Apurva Jain Cyber Security Manager

The Enterprise Immune System

Embed Size (px)

Citation preview

PowerPoint Presentation

The Enterprise Immune SystemMachine Learning for Cyber Threat DefenceApurva JainCyber Security Manager

PERSONAL INTRODUCTION

[introduce yourself]

Thank you for the opportunity to present to you.

Ill be talking to you about a new approach to cyber defense, known as the ENTERPRISE IMMUNE SYSTEM1

Company BackgroundFounded in 2013 in Cambridge, UKStarted by mathematicians and government intelligence specialistsTechnology based on machine learning & mathematicsHQs in Cambridge, UK & San Francisco Over 2000 customer installations23 global locations600% year-on-year growthDarktrace is a game-changer Virgin Trains

The Queens Awards for Enterprise Innovation 2016Bloomberg Business Innovator 2016Security Company of the Year at Info Security Global Excellence Awards 2016Best Insider Threat Detection and Solutions at Network Products Guide IT World AwardsGartner Cool Vendor 2015World Economic Forum Technology Pioneer 2015

VALIDATION

Founded in Cambridge by MATHEMATICIANS AND MACHINE LEARNING specialists, from the UNIVERSITY OF CAMBRIDGE

Over 750 CUSTOMER INSTALLATIONS across 20 COUNTRIES

Headquarters in Cambridge UK and San Francisco US [APAC: Singapore]

Widely recognized in the industry, by a variety of award bodies.

2

CUSTOMERS

Darktrace has customers across ALL INDUSTRY VERTICALS financial services, healthcare, retail, energy, transportation, non-profit etc.

Here are some of the organizations that we work with.

3

Evolving Threats in a New Business LandscapeOutsourced IT, SaaS, cloud, virtual, supply chainIts not just data breaches & defaced websitesEmergence of artificial intelligence attacks is leading to highly customised campaignsTrust attacks are silent and stealthyIntegrity of data is at riskInsider threat is constant whether malicious or non-malicious

Legacy controls are constantly outpaced

EVOLVING THREATS IN A NEW BUSINESS LANDSCAPE

Every day we are seeing attacks on the networks of our hundreds of customers around the world, and there are a couple of key trends which we see developing.

Firstly, the threat landscape is extremely challenging advanced and moving incredibly fast.

We read about data breaches and exfiltration in the news all the time. But it isnt these headline-making attacks that are the most dangerous.

Were starting to see much more complex, even machine learning based attacks, which use artificial intelligence to hide and move.

Secondly, we are seeing a rise in trust attacks attacks that are aimed at damaging a companys reputation or credibility by undermining the integrity of their data.

Often very stealthy, the attackers objective appears to be to change data slowly and over time.

They can remain in the network for hundreds of days before any damage is done.

In a bank or a hospital, these subtle changes over time can have a catastrophic effect on confidence in the data.

Insider threat is also a major concern. Anyone with access to your network is a potential insider from employees, to contractors, to customers and suppliers.

All this is against a backdrop of increasing business complexity, where the security industry is struggling to keep up with the threats posed by cloud and virtual environments, mobility, IoT, SaaS, and the simple fact of a network without traditional boundaries.

they have over a million devices.4

DARKTRACE APPROACH THE IMMUNE SYSTEM

Darktraces approach is different. This is a piece of DNA.

The DNA of our bodies is attacked all the time. Our bodies are continually confronted with new viruses and bacteria that must deal with.

Of course, skin protects us to a certain extent. But the critical piece of kit that we use for day-to-day is the IMMUNE SYSTEM the human immune system is one of the most complex systems in the biological world.

The immune system is intelligent, because it knows WHAT SELF IS in other words, it recognizes WHAT IS PART OF ME AND WHAT IS NOT PART OF ME and can identify ABNORMAL BEHAVIORS and take action against them. It is also changing and adapting to new environments.

Darktrace works in a similar way. Like the human immune system, it is a SELF-LEARNING SYSTEM, that CONTINUALLY EVOLVES AND ADAPTS to understand normal activity, and stop threats BEFORE THEY DEVELOP.The idea is not just to look for malicious actions but anomalies which could be as simple as may be a configuration problem but can develop into malicious activity if not addressed. There are lot of other things which can be very bad without it being malicious in nature. The EIS detects the subtle changes or movements as symptoms. Malaria symptom example

5

The Enterprise Immune System: Proven To WorkLearns self in real timeFor every individual user, device and network, using unsupervised machine learning

Finds the threats that get throughDetects both insider and sophisticated external threats, from within the network

100% visibilityVisualises entire network, including traditional and non-traditional IT, allows for investigations

ScalableLargest deployment has over 1 million users

All networks & devicesWorks on physical and virtual networks, cloud, ICS

ENTERPRISE IMMUNE SYSTEM

Darktrace technology is known as the Enterprise Immune System for its ability to replicate this self-learning capability of the immune system.

Its powered by machine learning and mathematics which uniquely model every user, device and the network as a whole.

Using these models, Darktrace determines an understanding of an organizations pattern of life.

This evolving pattern of life allows us to detect threats that get past legacy tools precisely because we are not relying on rules, signatures or assumptions of what bad is.

And it works in real time

Darktrace also provides 100% visibility of your network through our Threat Visualizer interface we will automatically visually map your network to give you a real-time overview of all user, device, network behaviours.

[And because we are correlating events and activity over time, you have a long view of network activity, and can replay historic events for investigation.]

Darktrace will also pick up things like IoT devices, activity in your cloud, and ICS environments, as well as your traditional IT

And it also scales up easilyour largest deployment is one of the top three banks in the world, and they have over a million devices.6

How Does Darktrace Work? Delivered as an appliance Passive tap into your networkAutomatically learns normal for all devices, users and the networkInterface accessed via web browserResults from day oneNo set-up

Installs in just 1 hour

INSTALLATION

PASSIVE TAP into your network

AUTOMATICALLY LEARNS nomal and abnormal behavior of your devices, users and network as a whole

RESULTS FROM DAY ONE

INSTALL IN ONE HOURIt doesnt rely on agents which saves the time otherwise required for installation and maintenance. And detects 20-30% more devoces than admins would expect7

How Does Darktrace Work? Delivered as an appliance Passive tap into your networkAutomatically learns normal for all devices, users and the networkInterface accessed via web browserResults from day oneNo set-up Installs in just 1 hour

INSTALLATION

PASSIVE TAP into your network

AUTOMATICALLY LEARNS nomal and abnormal behavior of your devices, users and network as a whole

RESULTS FROM DAY ONE

INSTALL IN ONE HOURIt doesnt rely on agents which saves the time otherwise required for installation and maintenance. And detects 20-30% more devoces than admins would expect8

Technology Architecture

ARCHITECTURE

So, what does the Darktrace immune system process look like?

Darktrace takes RAW NETWORK TRAFFIC

From that network data, we extract only RELEVANT METADATA using our proprietary DARKFLOW technique.

[Unlike sampling technologies, Darkflow infers and analyzes data from every packet captured, identifying crucial pieces of information needed for analyzing threats, e.g. initiators of machine connections, or the length of connections, etc.]

That data is then modelled, to understand the PATTERN OF LIFE for every USER, DEVICE AND NETWORK AS A WHOLE.

The THREAT CLASSIFIER does the job of classifying and FILTERING OUT FALSE POSITIVES, using Recursive Bayesian Estimation techniques.

THREAT NOTIFICATIONS AND NETWORK ACTIVITY are made visible to users via our interface, the THREAT VISUALIZER.

In addition, we also have a MODEL EDITOR, that allows you to create your own policies and rules, if required. Darktrace can also be fully INTEGRATED with existing outputs and dashboards.9

Machine Learning is Hard to Get Right

No two networks are alike needs to work in every networkOn-premise, virtualized, Cloud, SaaS, segmentedNeeds to work without customer configuration or tuning of modelsNeeds to support teams with varying security & maths skillsMust deliver value immediately, but keep learning and adapting as it goesMust have linear scalability Cannot rely on training sets of data

MATHEMATICS AND MACHINE LEARNING

Enterprise Immune System technology is underpinned by MACHINE LEARNING AND MATHEMATICS, which are fundamental to the difference of our approach.

Darktrace is based on a branch of BAYESIAN MATHEMATICS this is based around the idea of CORRELATING WEAK INFORMATION, in order to evaluate an outcome.

[this is Reverend Thomas Bayes, an English church cleric and mathematicians, who developed Bayes Theorem in an attempt to prove the existence of God]

RECURSIVE BAYESIAN ESTIMATION is a branch of Bayesian theory, developed by mathematicians from the UNIVERSITY OF CAMBRIDGE

It allows Darktrace to CONTINUALLY CALCULATE PROBABILITIES BASED ON EVOLVING EVIDENCE Darktraces understanding is constantly being revised and updated through new observations.

Critically, this approach DOES NOT REQUIRE PRIOR ASSUMPTIONS OR KNOWLEDGE of what is good or bad.

[further questions may be referred to a later conversation delighted to set up a call with our technical and mathematics teams if youd like more information]

MACHINE LEARNING IS HARD TO GET RIGHT

So, the fundamental basis of Darktraces technology is unsupervised machine learning.

The reality is that machine learning is very difficult to implement in practice.

There are several reasons for that:

Every network is different it needs to work across all types and sizes of network

A security solution cannot rely on sets of training data. [Networks are too complicated, and threats are constantly evolving.]

It has to scale.

It needs to deliver value immediately

What Darktrace has been able to do uniquely is to apply unsupervised machine learning in the real world.

Indeed, it has been proven time and time again to work across all types of network, with data moving at different frequencies, and it detects the threats that are quite simply going unnoticed by other approaches.

Critically our machine learning adapts, grows, and evolves with your business.

It is constantly calculating new probabilities based on evolving evidence.

[further questions may be referred to a later conversation delighted to set up a call with our technical and mathematics teams if youd like more information]

10

Immune System Technology Finds Threats That Go UndetectedOver 27,000 in-progress threats detected, including:Exfiltration of sensitive data by insidersHacked IoT devices, including HVAC, video conferencingThird-party contractor vulnerabilitiesPolymorphic & metamorphic malware that blend inIrregular VPN access from remote users & sitesCompromises of industrial control systemsIndiscriminate worms, Trojans, ransomwareAttacks on physical security, such as biometric scanners & badge readers

11

Compromise of Biometric Scanner

Industry: ManufacturingPoint of Entry: Fingerprint scannerApparent Objective: Alter biometric access keysGLOBAL THREAT CASE STUDY

Attacker successfully exploited known software vulnerabilities in fingerprint scannerAble to control information sent to and from the fingerprint scannerWent unnoticed by traditional anti-malware solutionsDarktrace detected unusual connections to and from the biometric scannerIf undetected, malicious actors would have gained access to physical machinery

COMPROMISE OF A BIOMETRIC (FINGERPRINT) SCANNER [Threat Anecdotes and Use Cases Script number 1]

So let me give you an example:

In this case, to protect its physical assets, a manufacturing company had installed biometric fingerprint scanners to access machinery.

In this instance, the close association of physical and network resources led to a hacker successfully exploiting a fingerprint scanner.

Not only did this attacker gain access to the fingerprint records stored by the system, but they were also able to add new records in order to gain unauthorized physical access to the company premises.

Now because this abnormal activity didnt correspond to any known attack signatures, traditional anti-malware solutions failed to detect the subtle and discrete operations that caused the compromise.

But Darktrace was able to detect the attackers movement using machine learning, because it had learnt what was normal behaviour for the scanner, and recognized that it had started behaving abnormally.

So, we were able to alert the company before any serious damage was caused.

12

Video Conferencing Camera HackVideo conferencing camera was transmitting data outside the networkCamera had been compromised by a remote attackerAttacker was aiming to either:Steal corporate informationTake remote control of the device to launch a DDoS attack on another networkWould not have been detected through signature-based defences the activity was not inherently malicious

Industry: RetailPoint of Entry: Video conference cameraApparent Objective: Transmit mass amounts of data out of host network

GLOBAL THREAT CASE STUDY

DATA EXFILTRATION FROM HACKED VIDEO CONFERENCING CAMERA [Threat Anecdotes and Use Cases Script number 3]

In this example, Darktrace found a hacked IoT device that was a dangerous vulnerability on a companys network

We detected unusual behaviour from a video conferencing camera in a companys network it was transmitting much larger volumes of data outside the network compared to similar devices.

The camera had been compromised by a remote attacker and was sending data possibly videos and photos outside of the network.

It was also connecting to other computers as the attacker explored the network and attempted to locate Point of Sale devices.

Darktrace detected this threat after the device initiated a very large upload to rare external IPs, and communicated with internal computers that it never usually talked to.

A back-door Trojan had been uploaded to the device before Darktrace was installed, allowing it be remotely accessed from outside the network.

The attacker was likely aiming to either:steal corporate information, including highly invasive audio and video feed data, or take remote control of the device to launch a DDoS attack on another network.

Either of these would have been a serious security risk to the company.

Once Darktrace flagged this behaviour, the company immediately disconnected the camera, and started a detailed review of their systems.

This would not have been detected through signature-based defences, as this activity was not inherently malicious.13

ConclusionThe battlefield is now inside corporate networks

Rules and signatures dont work

No security team, no matter how large, can keep up with new era of machine threats

Darktrace Enterprise Immune System is a fundamental new approach

30-day Proof of Value in your network

VALIDATION & CONFIDENCE this is what your audience should TAKE AWAY from the presentation

THE ENTERPRISE IMMUNE SYSTEM is a UNIQUE APPROACH POWERED BY MACHINE LEARNING & MATHEMATICS

IT UNDERSTANDS WHATS NORMAL FOR DEVICES, USERS AND THE NETWORK.

IT DETECTS THREATS AS THEY EMERGE

NO RULES, NO SIGNATURES

EASY TO INSTALL JUST ONE HOUR.

WE OFFER A PROOF OF VALUE (POV) where you can try the technology for a period of 4 weeks. 14

Thank you

Thank you for your time. [Discuss next steps]15