![Page 1: Rise of the Machine (Learning) - ANZTB · @Stefowilso - ANZTB 2017 MODES Learning vs Active - Learning = Imperva WAF is monitoring that URL and collecting information to …](https://reader030.vdocuments.mx/reader030/viewer/2022020302/5ada800c7f8b9a52528cec09/html5/thumbnails/1.jpg)
Rise of the Machine (Learning)
Stephanie Wilson
ANZTB - 2017@Stefowilso
![Page 2: Rise of the Machine (Learning) - ANZTB · @Stefowilso - ANZTB 2017 MODES Learning vs Active - Learning = Imperva WAF is monitoring that URL and collecting information to …](https://reader030.vdocuments.mx/reader030/viewer/2022020302/5ada800c7f8b9a52528cec09/html5/thumbnails/2.jpg)
@Stefowilso - CASTx17 2017
DON’T BE SCARED
![Page 3: Rise of the Machine (Learning) - ANZTB · @Stefowilso - ANZTB 2017 MODES Learning vs Active - Learning = Imperva WAF is monitoring that URL and collecting information to …](https://reader030.vdocuments.mx/reader030/viewer/2022020302/5ada800c7f8b9a52528cec09/html5/thumbnails/3.jpg)
@Stefowilso - CASTx17 2017
A I
MACHINE LEARNING
V R A R
CHATBOT
N L PN L U
DATA MINING
BIG DATA
NEURAL NETWORKS
DEEP LEARNING
PREDICTIVE ANALYTICS
DATA ANALYTICS
AND MORE...
![Page 4: Rise of the Machine (Learning) - ANZTB · @Stefowilso - ANZTB 2017 MODES Learning vs Active - Learning = Imperva WAF is monitoring that URL and collecting information to …](https://reader030.vdocuments.mx/reader030/viewer/2022020302/5ada800c7f8b9a52528cec09/html5/thumbnails/4.jpg)
@Stefowilso - ANZTB 2017
Artificial IntelligenceThe theory and development of computer systems able to perform tasks normally requiring human intelligence, such as visual perception, speech recognition, decision-making, and translation between languages.
Machine LearningMachine learning is a type of artificial intelligence (AI) that provides computers with the ability to learn without being explicitly programmed. Machine learning focuses on the development of computer programs that can change when exposed to new data
ChatbotA computer program designed to simulate conversation with human users
Natural Language ProcessingNatural language processing (NLP) is the ability of a computer program to understand human speech as it is spoken. NLP is a component of artificial intelligence (AI).
![Page 5: Rise of the Machine (Learning) - ANZTB · @Stefowilso - ANZTB 2017 MODES Learning vs Active - Learning = Imperva WAF is monitoring that URL and collecting information to …](https://reader030.vdocuments.mx/reader030/viewer/2022020302/5ada800c7f8b9a52528cec09/html5/thumbnails/5.jpg)
@Stefowilso - CASTx17 2017
FAUX MACHINE LEARNING
MACHINE LEARNING
CHATBOTS
THE PAST & FUTURE
![Page 6: Rise of the Machine (Learning) - ANZTB · @Stefowilso - ANZTB 2017 MODES Learning vs Active - Learning = Imperva WAF is monitoring that URL and collecting information to …](https://reader030.vdocuments.mx/reader030/viewer/2022020302/5ada800c7f8b9a52528cec09/html5/thumbnails/6.jpg)
@Stefowilso - ANZTB 2017
![Page 7: Rise of the Machine (Learning) - ANZTB · @Stefowilso - ANZTB 2017 MODES Learning vs Active - Learning = Imperva WAF is monitoring that URL and collecting information to …](https://reader030.vdocuments.mx/reader030/viewer/2022020302/5ada800c7f8b9a52528cec09/html5/thumbnails/7.jpg)
@Stefowilso - CASTx17 2017
WHAT DO YOU MEAN, YOU’RE A SETTING?
WHAT DO YOU MEAN, YOU’RE AUTOMATED?
![Page 8: Rise of the Machine (Learning) - ANZTB · @Stefowilso - ANZTB 2017 MODES Learning vs Active - Learning = Imperva WAF is monitoring that URL and collecting information to …](https://reader030.vdocuments.mx/reader030/viewer/2022020302/5ada800c7f8b9a52528cec09/html5/thumbnails/8.jpg)
@Stefowilso - CASTx17 2017
THE FIRST FORAY
![Page 9: Rise of the Machine (Learning) - ANZTB · @Stefowilso - ANZTB 2017 MODES Learning vs Active - Learning = Imperva WAF is monitoring that URL and collecting information to …](https://reader030.vdocuments.mx/reader030/viewer/2022020302/5ada800c7f8b9a52528cec09/html5/thumbnails/9.jpg)
@Stefowilso - ANZTB 2017
TESTING THE WAFWeb Application Firewall
go.xero.com
AWS
WAFChecking for:
Dodgy people
(Source IP, Geolocation)
Dodgy input/signatures
(HTTP headers, Cookies)
![Page 10: Rise of the Machine (Learning) - ANZTB · @Stefowilso - ANZTB 2017 MODES Learning vs Active - Learning = Imperva WAF is monitoring that URL and collecting information to …](https://reader030.vdocuments.mx/reader030/viewer/2022020302/5ada800c7f8b9a52528cec09/html5/thumbnails/10.jpg)
@Stefowilso - ANZTB 2017
MODESSimulation vs Active
- Simulation = Imperva WAF is monitoring traffic, and will give
us risk-based alerts based on the rules set, and will let the
traffic through
- Active = Imperva WAF will enforce rules and give us risk-
based alerts. High risk alerts with high probability (i.e. hardly
ever legit) are blocked.
![Page 11: Rise of the Machine (Learning) - ANZTB · @Stefowilso - ANZTB 2017 MODES Learning vs Active - Learning = Imperva WAF is monitoring that URL and collecting information to …](https://reader030.vdocuments.mx/reader030/viewer/2022020302/5ada800c7f8b9a52528cec09/html5/thumbnails/11.jpg)
@Stefowilso - ANZTB 2017
MODESLearning vs Active
- Learning = Imperva WAF is monitoring that URL and collecting
information to understand normal patterns (e.g. parameter
length used, methods used, etc). Traffic is let through.
- Active/Not learning = Imperva WAF creates rules based off the
behavior it has learned and throws low, medium, or high alerts
(i.e. blocks traffic) based off those rules.
![Page 12: Rise of the Machine (Learning) - ANZTB · @Stefowilso - ANZTB 2017 MODES Learning vs Active - Learning = Imperva WAF is monitoring that URL and collecting information to …](https://reader030.vdocuments.mx/reader030/viewer/2022020302/5ada800c7f8b9a52528cec09/html5/thumbnails/12.jpg)
@Stefowilso - CASTx17 2017
IT’S TESTING JIM, BUT NOT AS WE KNOW IT
![Page 13: Rise of the Machine (Learning) - ANZTB · @Stefowilso - ANZTB 2017 MODES Learning vs Active - Learning = Imperva WAF is monitoring that URL and collecting information to …](https://reader030.vdocuments.mx/reader030/viewer/2022020302/5ada800c7f8b9a52528cec09/html5/thumbnails/13.jpg)
@Stefowilso - CASTx17 2017
WHAT DID WE LEARN?
- Our team really disliked testing without a purpose
- Management didn’t really understand the terms exploratory and regression testing
- Our known testing techniques didn’t work
- It’s all about the data
![Page 14: Rise of the Machine (Learning) - ANZTB · @Stefowilso - ANZTB 2017 MODES Learning vs Active - Learning = Imperva WAF is monitoring that URL and collecting information to …](https://reader030.vdocuments.mx/reader030/viewer/2022020302/5ada800c7f8b9a52528cec09/html5/thumbnails/14.jpg)
@Stefowilso - CASTx17 2017
MACHINE LEARNING
MACHINELEARNING
![Page 15: Rise of the Machine (Learning) - ANZTB · @Stefowilso - ANZTB 2017 MODES Learning vs Active - Learning = Imperva WAF is monitoring that URL and collecting information to …](https://reader030.vdocuments.mx/reader030/viewer/2022020302/5ada800c7f8b9a52528cec09/html5/thumbnails/15.jpg)
@Stefowilso - ANZTB 2017
TYPES OF LEARNING
Supervised LearningCreating functions from known data sets, also known as labeled training dataTask Driven learning using Classification and Regression
Unsupervised LearningTraining Data no longer used and any manual labels are removedMachine is learning on it’s own using data driven clustering
Reinforcement LearningAlgorithm is learning to react to an environment on it’s own. There is a decision making process and reward system in place
![Page 16: Rise of the Machine (Learning) - ANZTB · @Stefowilso - ANZTB 2017 MODES Learning vs Active - Learning = Imperva WAF is monitoring that URL and collecting information to …](https://reader030.vdocuments.mx/reader030/viewer/2022020302/5ada800c7f8b9a52528cec09/html5/thumbnails/16.jpg)
@Stefowilso - CASTx17 2017
![Page 17: Rise of the Machine (Learning) - ANZTB · @Stefowilso - ANZTB 2017 MODES Learning vs Active - Learning = Imperva WAF is monitoring that URL and collecting information to …](https://reader030.vdocuments.mx/reader030/viewer/2022020302/5ada800c7f8b9a52528cec09/html5/thumbnails/17.jpg)
@Stefowilso - CASTx17 2017
![Page 18: Rise of the Machine (Learning) - ANZTB · @Stefowilso - ANZTB 2017 MODES Learning vs Active - Learning = Imperva WAF is monitoring that URL and collecting information to …](https://reader030.vdocuments.mx/reader030/viewer/2022020302/5ada800c7f8b9a52528cec09/html5/thumbnails/18.jpg)
@Stefowilso - CASTx17 2017
COULD WE DO MORE?
![Page 19: Rise of the Machine (Learning) - ANZTB · @Stefowilso - ANZTB 2017 MODES Learning vs Active - Learning = Imperva WAF is monitoring that URL and collecting information to …](https://reader030.vdocuments.mx/reader030/viewer/2022020302/5ada800c7f8b9a52528cec09/html5/thumbnails/19.jpg)
@Stefowilso - ANZTB 2017
INVOICING WITH MACHINE LEARNING?
![Page 20: Rise of the Machine (Learning) - ANZTB · @Stefowilso - ANZTB 2017 MODES Learning vs Active - Learning = Imperva WAF is monitoring that URL and collecting information to …](https://reader030.vdocuments.mx/reader030/viewer/2022020302/5ada800c7f8b9a52528cec09/html5/thumbnails/20.jpg)
@Stefowilso - ANZTB 2017
SIMPLE RULES VS MACHINE LEARNING
Would simple rules be good enough?
- Last used account code
- Most frequently used
- Default sales code
What Models could we use?
- Multinomial Naive Bayes
- Logistic regression
Invoicing Simulator in Python
- Common way to compare methods
- F1 scores (1 is perfect, 0 is opposite)
- 2000 Organisations + 2 accounts
- 150 Invoices
- How much data do we need?
![Page 21: Rise of the Machine (Learning) - ANZTB · @Stefowilso - ANZTB 2017 MODES Learning vs Active - Learning = Imperva WAF is monitoring that URL and collecting information to …](https://reader030.vdocuments.mx/reader030/viewer/2022020302/5ada800c7f8b9a52528cec09/html5/thumbnails/21.jpg)
@Stefowilso - ANZTB 2017
THE SET UP
![Page 22: Rise of the Machine (Learning) - ANZTB · @Stefowilso - ANZTB 2017 MODES Learning vs Active - Learning = Imperva WAF is monitoring that URL and collecting information to …](https://reader030.vdocuments.mx/reader030/viewer/2022020302/5ada800c7f8b9a52528cec09/html5/thumbnails/22.jpg)
@Stefowilso - ANZTB 2017
TRAINING MODEL
Credit - Brad Stillwell
![Page 23: Rise of the Machine (Learning) - ANZTB · @Stefowilso - ANZTB 2017 MODES Learning vs Active - Learning = Imperva WAF is monitoring that URL and collecting information to …](https://reader030.vdocuments.mx/reader030/viewer/2022020302/5ada800c7f8b9a52528cec09/html5/thumbnails/23.jpg)
@Stefowilso - ANZTB 2017
RESULTS
![Page 24: Rise of the Machine (Learning) - ANZTB · @Stefowilso - ANZTB 2017 MODES Learning vs Active - Learning = Imperva WAF is monitoring that URL and collecting information to …](https://reader030.vdocuments.mx/reader030/viewer/2022020302/5ada800c7f8b9a52528cec09/html5/thumbnails/24.jpg)
@Stefowilso - CASTx17 2017
WHERE DOES THIS LEAVE US?
![Page 25: Rise of the Machine (Learning) - ANZTB · @Stefowilso - ANZTB 2017 MODES Learning vs Active - Learning = Imperva WAF is monitoring that URL and collecting information to …](https://reader030.vdocuments.mx/reader030/viewer/2022020302/5ada800c7f8b9a52528cec09/html5/thumbnails/25.jpg)
@Stefowilso - ANZTB 2017
WELLINGTON CoP MISSION STATEMENT
"We aim to deliver experiences that delight customers by influencing the Xero culture to be more quality
oriented and transforming our software from 'good' to 'wow'!"
![Page 26: Rise of the Machine (Learning) - ANZTB · @Stefowilso - ANZTB 2017 MODES Learning vs Active - Learning = Imperva WAF is monitoring that URL and collecting information to …](https://reader030.vdocuments.mx/reader030/viewer/2022020302/5ada800c7f8b9a52528cec09/html5/thumbnails/26.jpg)
@Stefowilso - CASTx17 2017
IT’S TESTING JIM, BUT NOT AS WE KNOW IT
![Page 27: Rise of the Machine (Learning) - ANZTB · @Stefowilso - ANZTB 2017 MODES Learning vs Active - Learning = Imperva WAF is monitoring that URL and collecting information to …](https://reader030.vdocuments.mx/reader030/viewer/2022020302/5ada800c7f8b9a52528cec09/html5/thumbnails/27.jpg)
@Stefowilso - ANZTB 2017
THERE’S A LOT TO DO!
ANALYSIS- What is a good candidate for machine learning?- What does success look like? How do you measure it?- What is a good data set for supervised learning?- Do you need to A/B test? - How do you move into production? Do you need a target customer type?
BE CURIOUS- This is a young field, learn about it and what other people are doing- How can this new technology help you to do your job?- How can ML and AI be applied to new features or designs?- Can you improve the UX by reducing exceptions or possibility of customer error?
![Page 28: Rise of the Machine (Learning) - ANZTB · @Stefowilso - ANZTB 2017 MODES Learning vs Active - Learning = Imperva WAF is monitoring that URL and collecting information to …](https://reader030.vdocuments.mx/reader030/viewer/2022020302/5ada800c7f8b9a52528cec09/html5/thumbnails/28.jpg)
@Stefowilso - ANZTB 2017
IF YOU WANT TO LEARN MORE...
![Page 29: Rise of the Machine (Learning) - ANZTB · @Stefowilso - ANZTB 2017 MODES Learning vs Active - Learning = Imperva WAF is monitoring that URL and collecting information to …](https://reader030.vdocuments.mx/reader030/viewer/2022020302/5ada800c7f8b9a52528cec09/html5/thumbnails/29.jpg)
LOOK EVERYWHERE
@Stefowilso - ANZTB 2017
![Page 30: Rise of the Machine (Learning) - ANZTB · @Stefowilso - ANZTB 2017 MODES Learning vs Active - Learning = Imperva WAF is monitoring that URL and collecting information to …](https://reader030.vdocuments.mx/reader030/viewer/2022020302/5ada800c7f8b9a52528cec09/html5/thumbnails/30.jpg)
@Stefowilso - CASTx17 2017HOW CAN I HELP?
CHATBOTS
![Page 31: Rise of the Machine (Learning) - ANZTB · @Stefowilso - ANZTB 2017 MODES Learning vs Active - Learning = Imperva WAF is monitoring that URL and collecting information to …](https://reader030.vdocuments.mx/reader030/viewer/2022020302/5ada800c7f8b9a52528cec09/html5/thumbnails/31.jpg)
@Stefowilso - ANZTB 2017
CHATBOT LOLZ
![Page 32: Rise of the Machine (Learning) - ANZTB · @Stefowilso - ANZTB 2017 MODES Learning vs Active - Learning = Imperva WAF is monitoring that URL and collecting information to …](https://reader030.vdocuments.mx/reader030/viewer/2022020302/5ada800c7f8b9a52528cec09/html5/thumbnails/32.jpg)
@Stefowilso - ANZTB 2017
HELPFUL CHATBOTS
![Page 33: Rise of the Machine (Learning) - ANZTB · @Stefowilso - ANZTB 2017 MODES Learning vs Active - Learning = Imperva WAF is monitoring that URL and collecting information to …](https://reader030.vdocuments.mx/reader030/viewer/2022020302/5ada800c7f8b9a52528cec09/html5/thumbnails/33.jpg)
@Stefowilso - ANZTB 2017
![Page 34: Rise of the Machine (Learning) - ANZTB · @Stefowilso - ANZTB 2017 MODES Learning vs Active - Learning = Imperva WAF is monitoring that URL and collecting information to …](https://reader030.vdocuments.mx/reader030/viewer/2022020302/5ada800c7f8b9a52528cec09/html5/thumbnails/34.jpg)
@Stefowilso - CASTx17 2017
INDUSTRIALINTELLIGENCE REVOLUTION
![Page 35: Rise of the Machine (Learning) - ANZTB · @Stefowilso - ANZTB 2017 MODES Learning vs Active - Learning = Imperva WAF is monitoring that URL and collecting information to …](https://reader030.vdocuments.mx/reader030/viewer/2022020302/5ada800c7f8b9a52528cec09/html5/thumbnails/35.jpg)
@Stefowilso - CASTx17 2017
HUMANS ARE INEFFICIENT
HUMANS DO THINGS DIFFERENTLY
![Page 36: Rise of the Machine (Learning) - ANZTB · @Stefowilso - ANZTB 2017 MODES Learning vs Active - Learning = Imperva WAF is monitoring that URL and collecting information to …](https://reader030.vdocuments.mx/reader030/viewer/2022020302/5ada800c7f8b9a52528cec09/html5/thumbnails/36.jpg)
@Stefowilso - CASTx17 2017
DON’T BE SCARED
![Page 37: Rise of the Machine (Learning) - ANZTB · @Stefowilso - ANZTB 2017 MODES Learning vs Active - Learning = Imperva WAF is monitoring that URL and collecting information to …](https://reader030.vdocuments.mx/reader030/viewer/2022020302/5ada800c7f8b9a52528cec09/html5/thumbnails/37.jpg)
Thank you!@Stefowilso
Stephanie Wilson
ANZTB 2017