By Julian Arango | May 06, 2019
Imagine a medium-sized sealed carton box, with two or three glasses inside. If you kick the box (like kicking a soccer ball), the glasses will surely break. The glasses are fragile. Now, think of the same box, but with two or three standard steel hammers. Nothing will happen to those hammers after kicking the box. The hammers are robust in this context.
In cybersecurity, there are plenty of instances
in which artifacts or entities are fragile
given some inputs (we will call them stressors).
For example, some default password designs are fragile.
It is cheap to find lists of these and it is not so challenging
to perform a brute-force attack.
At some point, that stressor will break at least one security layer.
On the other hand, adding more security layers to authentication
(i.e., two-factor authentication) prevents systems to be hacked so easily;
these are robust (although how robust is depends on the type of stressor).
The typical paradigm of information security has been to go
from fragile designs and processes
to robust ones as previously unthinkable events appear.
Think of how networking protocols have evolved or how
is becoming more script-oriented
or transitioning into infrastructure as a code.
A couple of years ago, a new related concept appeared: antifragility. This term is developed in the book Antifragile by Nassim Taleb. A controversial figure, Taleb is a professor of Risk Engineering at New York University. Beyond robustness, he points to actions and entities that, rather than protect, gain from randomness, disorder, and uncertainty. He called this antifragile. Lifting weights and administering vaccines are antifragile actions: people become stronger from this stressors. Lifting weights break muscle fibers. Vaccines are small illnesses injected. These make your body to grow stronger muscles and develop resistance against viruses, respectively. Antifragility is becoming better from struggles. If you google a bit, you might find that a Hydra, the mythological monster, is an antifragile entity.
Shane Parrish, who runs Farnam Street blog,
has provided more examples related to this concept using a triad:
Here are two cases:
Mistakes are just information
Loves mistakes (since they are small)
Studying events, measuring their risks, statistical properties of events
Studying exposure to events, statistical properties of exposures
Modifying exposure to events
How can we foster antifragility in our cybersecurity efforts? We will discuss two ways: first, by creating “troubles” and inject them into operations. Second, by experimenting, tinkering and exposing ourselves and systems to small risks.
The worst nightmare for a company like
Netflix is downtime.
In 2008, the company decided to migrate to a cloud infrastructure
after facing a massive database incident.
Complete migration took eight years (read here about it).
In the process, back in 2010,
the company launched
a tool to cause their cloud servers to be offline by random.
They cannot afford, for instance, outages of their
“By running Chaos Monkey in the middle of a business day, in a carefully monitored environment with engineers standing by to address any problems, we can still learn the lessons about the weaknesses of our system, and build automatic recovery mechanisms to deal with them. So next time an instance fails at 3 am on a Sunday, we won’t even notice.”
By being exposed to these outages,
Netflix assesses how proper the countermeasures are to keep services running.
If they are not, they devise improvements
to avoid these outages when they happen for real.
“By pseudo-randomly rebooting their own hosts, they could suss out any weaknesses and validate that their automated remediation worked correctly,”
a “chaos engineering” company.
Netflix didn’t stop there.
They build a
more tools to inject other “troubles” into their platform.
One of these is
Chaos Gorilla which simulates an outage
of an entire availability zone.
This mindset is in line with the cases shown in
Netflix was not waiting for mistakes to happen; they created them.
Furthermore, those "mistakes" were diverse;
they were creating different exposures to their platform.
The journey has not been perfect,
and some outages took place,
especially at the early days of the migration
(read about a Christmas outage here).
However, Netflix has achieved a remarkable resiliency,
even with underlying infrastructure failing
(see here a report on Amazon S3 outage),
making those earlier outages something tiny.
I don’t know about you, dear reader,
but I’ve never experienced a
Netflix outage since July 2013,
when I became a customer.
“The test of all knowledge is experiment. Experiment is the sole judge of scientific ‘truth’.” - Richard Feynman
In my work, I rely on behavioral insights most of the time to think about behavior change. Many people I have worked with believe Behavioral Science is the origin of proper experimentation, but they are wrong. It is precisely the opposite: good experimentation has been essential in creating what Behavioral Science is today. Likewise, good experimentation is a door-opener for novel ideas, different approaches in doing things regardless of discipline. Some people think innovations come from a bunch of new methodologies and frameworks (e.g., design thinking), but I would argue that innovations are created more from an exploration mindset and curiosity than anything else. The glue of these two elements is running experiments. By exploring and being curious, we discover how the world works; we start creating hypotheses. By doing experimentation, we test those hypotheses and continue to build knowledge upon empirical results.
In a sense, becoming antifragile means that you would have to crave for some chaos. You would have to seek for variability, and you would have to embrace and face uncertainty. This would lead you to learn to live in harsh circumstances if they arrive. To become antifragile is to become a constant learner with diverse inputs. In this way, a company could be always a step further and be better prepared for the uncertain future. That’s exactly what experiments are for: to learn what works and what doesn’t.
I acknowledge that the corporate world has reasons not to invest or support this mindset. One of these is the cost. Investing in experiments, with a no clear outcome isn’t encouraging. However, in the words of Parrish, that’s playing the short-term game, which is dangerous. Another reason is how to handle results from unsuccessful experiments. People in firms fear they would seem ridiculous in the face of unfavorable results. But, here’s the thing: how much money companies are not capturing by not experimenting? It might be they are just following the herd. We should reframe our thinking into asking how much are we missing (or losing) by not experimenting, rather than keeping telling ourselves that an experiment is expensive.
“You have to be willing to look like an idiot in the short term to look like a genius in the long term” - Shane Parrish —Farnam Street blog
John List and Uri Gneezy (both experimental economists),
in their book The Why Axis,
suggest that businesses that do not experiment are losing money.
Furthermore, they claim that executives
of these non-experiment-oriented corporations
will become endangered species (
An organization I admire is The Behavioral Insights Team (
This company works designing and running experiments since 2010.
Almost every project is a well-designed experiment.
They aim to find cheaper and scalable solutions
to improve efficiency and efficacy of services
from governments and public institutions towards citizens.
They combine novel approaches including behavioral science,
adequate randomized control trials, and data analyses.
From their work (and learning),
they were able to launch
their product development army.
They have been looking for "mistakes"
and experimenting with different “exposures”
to help governments stepping away from fragility.
In their journey,
BIT itself has become a great example
of an antifragile company.
BIT has been featured:
Want to start becoming antifragile in your cybersecurity efforts?
We can help.
Check out our services.
We inject a mix of “troubles” into your applications and
We are able to create some monkeys
and even gorillas to shake your
making offensive testing your best line of protection.
That way, you can learn how to better prepare for potential incidents
and outthink attackers!
And don’t worry: we do it in a controlled way.
We also invite you to take a look at Integrates, our platform to keep track of weaknesses. Think of it as a platform that helps you to learn how to become antifragile in cybersecurity.
Corporate member of The OWASP Foundation