EnterpriseImmuneSystem

The Enterprise Immune System
WHITE PAPER
Embracing Probability to Deliver
Next-Generation Cyber Defense

2
Executive Summary
Darktrace takes a fresh approach to the challenge of
countering sophisticated attackers coming from the
outside or the inside of an organization, based on
advanced machine learning and mathematics.
If we look back to medieval times, the concept of a
walled hill town was common, whereby the people
inside the town walls were trusted and the people on
the outside were not. This model was reassuring and
worked well, until the interests of growth and trade
made it necessary for all sorts of individuals to interact
with each other. The modern city thus came into being
based on dynamic collaboration and interaction – and
no city walls in sight. The most modern and successful
cities in the world are complex places and continue mile
after mile.
fundamental to progress, and we have seen this in the
cyber security area too. Where walls are put up, they
will often be overcome, not only by malicious attackers
but also by company employees who choose to bypass
As we consider risk assessment, it is important to
remember that your employees will not always take the
path that you have chosen for them.
Data is not only at risk of being illegally taken out of
the organization, but also being unknowingly changed.
Whereas data loss or theft loss is undesirable, it is
usually obvious when it has happened. Much more
of data has been compromised. A high premium is put
what happens when 10,000 bank account balances are
changed, but the bank is unable to identify which ones?
A new approach that pays attention to very subtle
actions that happen very quietly will be critical to
winning the battle against cyber attackers. It is a battle
that has been played out in the biological world too for
billions of years. Our DNA represents information that is
constantly under attack from a variety of viruses. Whilst
we cannot live in plastic bubbles and avoid these viral
attacks, we have an immune system that understands
what ‘self’ is, knowing what is part of ‘us’ and therefore
identifying what is abnormal and not ‘us’.
Dealing with the unexpected in this way is crucial in
our ability, as human beings, to protect against major
threats to our health while interacting and collaborating
with each other. Darktrace helps companies strike
this balance by delivering an immune system for the
enterprise that calculates probabilities of threats in the
light of evidence. Based on proven machine learning
and mathematics, it allows organizations to stay close
to their employees, supply chain and customers, whilst
protecting against serious, sophisticated threats. This
deal with our new cyber reality, when we don’t always
The Threat Is Inside
While the traditional approach to security relies on a
distinction between inside and outside, the reality is
modern infrastructures, and threat actors are already on
the inside. Today’s information systems and networks
are global, complex and porous – they need to be in
order to work and prosper in the modern world. But this
makes them vulnerable to exploitation too.
Today, all companies face risk from within their own
walls from a variety of individuals and groups, including
their employees and contractors, supply chain and
customers, as well as malicious parties that have
transpire week by week, often hitting the headlines
and entailing serious damage to the victim company’s
reputation and revenues, show that traditional methods
are not working.
It is no good building a wall around a system that is
with which the modern business is expected to act.
Companies are in danger of suffering from the double
negative impact of damaging their competitiveness and
productivity, due to overzealous data restrictions, and at
the same time remaining extremely vulnerable to both
insider threats and sophisticated external attackers.
The real ‘cyber security’ challenge is not about ‘securing’
our information systems – it is about accepting
probability and understanding what is really going on
from the inside of complex information environments.
It is an illusion to think that if we have a lock and a key,
we can achieve perfect security. In reality, we know that
locks break and keys can be copied – and there is no

White Paper
3
“Good cyber security is not juﬆ
about a really ﬆrong wall on
the outside, but some kind of an
immune syﬆem within.”
MI5
perfect answer. The best solution must allow companies
to continue to expose themselves to all sorts of risks,
in the interests of running global, connected and
competitive businesses.
Whilst this approach may protect against
unsophisticated attackers that use the same toolkit
and tactics repeatedly, it is widely known that the most
serious attackers are very good at what they do – they
change their strategies all the time and use bespoke
Furthermore, for this rules-based approach to work to
any level of accuracy, constant software updates are
required, although inevitably fail to keep up with the
pace of constantly-evolving attack tactics. It is reactive
and incapable of defending against fresh, ever-changing
threats as they happen.
3. The threat is not purely technical
It can be easy to forget the human at the end of every
cyber mission, in a world of botnets, Trojans and Remote
Access Tools. It is important to remember that serious
cyber threats are directed by skilled people who move
deftly and subtly around the network. The traditional
black and white approach described above is incapable
of dealing with the complexity and delicacy that such
attackers bring to their missions.
Both external and insider parties usually exhibit
engaging in malicious acts. A contractor logging into the
aggregated, the set-up of a new employee account or
but form a compelling picture when correlated.
In addition to the human nature of the attack mission,
we should remember that the human vulnerability of
employees and insiders is often used and exploited by
and aim to dupe innocent-minded insiders into taking a
certain action. Whilst training can help reduce this, it is
impossible to stop all your employees taking the wrong
decision every time.
The challenge of cyber and IT security executives
is to embrace probability, taking into account the
subtleties of human behaviors. This requires seeing
and understanding subtle, weak indicators that point
to human intent and activity from within the network,
in order to detect the emerging threat behaviors of the
astute individuals behind every potential threat.
Wrong Assumptions: Why
the Traditional Approach Is
Failing
There are three key reasons why the traditional IT
security approach to cyber security has failed.
1. You can’t keep threat out
The traditional approach to security assumes that you
can keep attackers on the outside of your networks,
by strengthening your boundaries. Organizations
have invested large amounts of time and money into
the goal of keeping their information systems free from
Unfortunately, the majority of corporate networks have
proven themselves capable of overcoming perimeter
controls, and employees and other insiders with access
work on the basis that we are constantly at risk and that
many threats, and certainly the most insidious, will get
into an organization with relative ease.
2. You can’t deﬁne what is illegitimate
information technology. This core IT principle, widely
used for all manner of automated applications, has duly
there is a wide range of solutions that aim to create
protect you against that type of ‘bad’, if and when it is
encountered again, in exactly the same way.

4
Protecting Data Integrity – It’s
Not Juﬆ About Data Loss
The nature of risk has changed, broadening to
encompass a whole range of ways in which damage can
loss and theft, with the proliferation of individuals and
groups of individuals who want to take data because of
This is a worrying phenomenon, but perhaps more
concerning still is that today’s information environments
allow for data to be changed, without this being
detected. Whether you are an insider or an external
hacker, once you are on the inside, your access to data
and your ability to change that data, or parts of it, is
serious problems can arise. Imagine a situation where
even your back-up is corrupted.
Ensuring data integrity is absolutely central for most
businesses in order to operate in the public market. For
a healthcare organization handling patient blood types
or a bank managing its customers’ bank balances, the
prospect that critical data could be compromised in some
their products or services and put them out of business.
When we think about how to solve the problem of cyber-
attacks, we need to think beyond isolated data leaks
or website hacks and take in the bigger implications of
defended and most security-conscious organizations
are vulnerable to subtle, lone attackers who move
silently within their systems and have the ready means
to disproportionately undermine their entire operation.
Insider threat
The threat from people inside the organization is
massively underestimated – perhaps because it is an
our employees and partners with certain access rights
to allow them to do their jobs, and yet at the same time,
we cannot trust all of them to take the right decision,
and unwittingly help an attacker get in; others may have
a grudge or other motivations that drive them to abuse
their privileges with malicious intent.
Insider threat does not only come from our employees,
but anyone with access to the network, corporate
data or company premises, which makes policing
insiders challenging. Modern enterprises thrive on
the interconnectedness between them and their
supply chain, customer base, contractual workers and
other partner organizations and individuals. Today’s
corporations have expanded their networks accordingly,
opening up their organizations to increased risk for
operations.
There is a wide range of reasons why an insider could be
inspired to act, either unwittingly or with deliberate intent,
loyalty to family or friends and general disaffection. With
a range of personalities and motivations involved, it is
impossible to identify high-risk users in advance.
Rather than analyze ‘who did it’ or try to second guess
‘who might do it’, the real challenge is to identify such
attacks and understand how they develop in real
time – by analyzing and correlating the subtle signs
of compromise that each insider makes as they move
within the network.
Insider incidents do not need to be at the scale of
damage to your reputation or business. In spite of
the improvements to internal security in the majority
of enterprises, we know that insider attacks are still
feasible to carry out in spite of legacy security tools and
good practise. Indeed, as network boundaries become
and outside the organization has lost its relevance. One
way or another, today’s threats are carrying out the bulk
learn and interpret the behaviors of all users, devices
and networks in order to spot real adversaries in a world
in which everyone is – or can become – an insider.
“A key element of
improvement involves
acknowledging the importance
of human behavior when
designing, building and using
cyber security technology.”
Institute for Information Infrastructure
Protection, 2013

White Paper
5
“The key is detecting that you
have been breached as quickly
as possible and having the
syﬆems in place to protect
your vital assets.”
2013
Introducing the Enterprise
Immune Syﬆem
Darktrace has pioneered a fundamentally new approach
to the cyber challenge, known as the Enterprise
that organizations face a constant level of threat from
within. This emerging category represents cutting-edge
technology that is capable of learning ‘self’ within an
organization on an adaptive, real-time basis – thereby
understanding when abnormal behavior starts to
manifest itself.
Like viral DNA, which constantly mutates and evolves
to ensure its survival within the human body, cyber
attackers are sophisticated and constantly change
and tweak their behaviors in order to avoid detection.
Fortunately for us, the immune system is just as clever as
viral DNA – it is continually learning and understanding
what constitutes a threat. It’s not a perfect system –
sometimes we catch the odd cold – but generally it does
a great job at protecting us from serious illnesses. In
doing so, it allows humans to interact with each other
and expose themselves to risk on a day-to-day basis.
Because human beings thrive on social interaction and
collaboration, living in a sterile glass box is not an option
for us – and it is not an option for modern enterprises
either.
Darktrace’s innovative approach is based on complex
mathematics that calculates probabilities in the light
of evidence. This probabilistic method of detecting
anomalies is the most pragmatic and accurate way of
protecting against unknown attack vectors operating
within complex network environments. Enterprise
of life for every network, device and individual user,
correlating this information in order to establish an
overview pattern of life and thereby spot deviations that
indicate live, in-progress threats.
This new approach does not require prior knowledge
of threats, looking instead for any behaviors that are
probabilistically determined to be anomalous and
therefore worthy of investigation. An Enterprise Immune
suspicious activity before major damage is done. It
also allows threat analysts to focus on truly concerning
incidents that are likely to be seriously threatening,
This ability to self-learn and adapt to changing
environments in real time represents a step change
for organizations around the world, enabling them to
reconcile their need for an interconnected workforce,
customer base and supply chain, whilst ensuring that
they protect against serious, existential threats to their
businesses in the most effective way possible.
Mathematics and Machine
Learning ‘Done Right’
The core of Darktrace’s innovative approach lies in a
breakthrough in probabilistic mathematics made at the
University of Cambridge. Bayesian theory is known
for its ability to draw meaning from large sets of data,
development of Darktrace’s founding technological
innovation.
By mathematically characterizing what constitutes
‘normal’ behavior, based on the analysis of multiple
data sources, RBE mathematics succeeds in identifying
changing attack behaviors where conventional
signature-based methods fall down. Powered by
RBE, Darktrace’s mathematical models are constantly
adapting themselves in real time, according to the new
information that it processes, and continually providing
calculations of threat levels.
probability state variable. This distribution is built from
observations or ‘features’. These features are recorded
iteratively and processed in real time on the platform.

6
A plausible representation of the relational information
among entities in dynamic systems in general, such as
an enterprise network, a living cell or a social community,
or indeed the entire internet, is a stochastic network,
which is topologically rewiring and semantically evolving
over time.
In many high-dimensional structured I/O problems, such
within an enterprise LAN or WAN, where both input and
output can contain tens of thousands, sometimes even
learning a sparse and consistent structured predictive
function is challenged by a lack of normal distribution.
In this context, Darktrace has pioneered the most
advanced, large-scale computational approach to learn
sparse structured I/O models by extending the L1-
family of sparse ‘structured’ regression models. This
allows for the discovery of true associations between
convex optimization problems and yield parsimonious
models.
Acute methods for estimating and analyzing varying
unknown times or locations are required for Enterprise
are frequently encountered in social and biological
problems, where data is structured and longitudinal,
and the IID assumptions on samples being generated
from an invariant underlying model no longer hold.
For example, at a given time point, the observations
used for estimating models corresponding to other time
points.
Darktrace has pioneered Bayesian methods for tracking
the changing model structures and parameters,
incorporating structural changes, the change times
and unknown variables. These methods are essential
when observing subtle variations in machine events to
determine pivotal features within a behavioral history
that may determine compromise.
In addition, the new mathematics provides Darktrace
with a non-frequentist architecture for inferring and
testing causal links between explanatory variables,
Bayesian belief networks and the new approaches
drawn, without the need for protracted and repeated
observation.
The core of Darktrace's mathematical processing is the
determination of normative behavior using the methods
described above, pivoting on Recursive Bayesian
techniques. This core incorporates a series of adaptive
“Darktrace is intereﬆing
because of its back-to-front
approach to security... [it]
proﬁles not possible attack
vectors, but the network itself,
as well as the devices that
connect to the network and the
network's users.”

White Paper
7
Conclusion
In an era of pervasive internal threat, a new approach
is necessary that does not rely on legacy distinctions
of ‘inside’ and ‘outside’ but instead is capable of
understanding what is happening from within an
organization and dealing with the unexpected.
Modern organizations are not like medieval walled cities
– interaction and collaboration across geographical and
virtual boundaries is vital to their ability to survive. The
new model of security must allow and encourage this
that the battle against it will be constant.
organizations to understand threat holistically and
defend against it, based on new machine learning
and mathematical breakthroughs. The system is
continually learning, meaning that it can deal with the
unpredictability of sophisticated threats and move in
step with both a changing organizational environment
and evolving threat landscape.
Organizations that have implemented an Enterprise
advances in machine learning and mathematics to
protect against insidious and persistent threats from
and interconnection that we all thrive on. An Enterprise
accepts the complexity of our systems and the threats
within them. You may still catch the odd cold but in this
game of probability, it’ll be a price well worth paying.
“Tackling cyber security
properly will require a leap of
faith into the more uncertain
world of probability – we need
to ﬆart picking our battles and
deal with the important ﬆuﬀ
that might kill us ﬁrﬆ.”
Dr Mike Lynch OBE, founder of Autonomy

ABOUT DARKTRACE
Darktrace is one of the world’s fastest-growing cyber defense companies and the leader in
Bayesian mathematics developed at the University of Cambridge. Darktrace addresses the
challenge of insider threat and advanced cyber-attacks through its unique ability to detect previously
learning platform to spot anomalous activity within the enterprise, in sectors including energy and
by world-class machine learning specialists and operational government intelligence experts. The
Francisco and Washington D.C.
CONTACT US
Email: info@darktrace.com
www.darktrace.com

EnterpriseImmuneSystem

More Related Content

What's hot (20)

Viewers also liked (13)

Similar to EnterpriseImmuneSystem (20)

EnterpriseImmuneSystem