2. What we will cover in this session
• Fraud Detection
• Machine Learning @ Amazon
• Amazon SageMaker
• Using SageMaker to Detect Fraud
3. Payment Fraud is an ongoing concern for FS organizations
In 2016…
$9b of fraud losses
in the US*
$22.8b of fraud
losses globally*
In 2021…
$32.96b of projected
fraud losses globally*
* From The Nilson Report, October 2017, Issue 1118 (https://nilsonreport.com/upload/content_promo/The_Nilson_Report_Issue_1118.pdf)
4. Types of Fraudulent Behaviour
Endpoint Authentication
☞ stolen card or machine
Layer 1
5. Endpoint Authentication
☞ stolen card or machine
Layer 1
Anomaly within a session
☞ Irregular behaviour within a session—e.g. transfer before balance
Layer 2
Types of Fraudulent Behaviour
6. Endpoint Authentication
☞ stolen card or machine
Layer 1
Anomaly within a session
☞ Irregular behaviour within a session—e.g. transfer before balance
Layer 2
Anomaly within an account
☞ Irregular transactions—e.g. spike in transfer and recipients
Layer 3
Types of Fraudulent Behaviour
7. Endpoint Authentication
☞ stolen card or machine
Layer 1
Anomaly within a session
☞ Irregular behaviour within a session—e.g. transfer before balance
Layer 2
Anomaly within an account
☞ Irregular transactions—e.g. spike in transfer and recipients
Layer 3
Anomaly within multiple channels of the same account
☞ Irregular transactions across channels—e.g. spike in transfer and recipients
Layer 4
Types of Fraudulent Behaviour
8. Endpoint Authentication
☞ stolen card or machine
Layer 1
Anomaly within a session
☞ Irregular behaviour within a session—e.g. transfer before balance
Layer 2
Anomaly within an account
☞ Irregular transactions—e.g. spike in transfer and recipients
Layer 3
Anomaly within multiple channels of the same account
☞ Irregular transactions across channels—e.g. spike in transfer and recipients
Layer 4
Anomaly within multiple channels of multiple accounts
☞ Irregular transactions across channels and accounts
Layer 5
Types of Fraudulent Behaviour
19. The AWS Machine Learning Stack
PLATFORMS
APPLICATION SERVICES
R E K O G N I T I O N R E K O G N I T I O N
V I D E O
P O L L Y T R A N S C R I B E T R A N S L A T E C O M P R E H E N D L E X
Amazon SageMaker Amazon Mechanical Turk
FRAMEWORKS KERAS
P3
NVIDIA Tesla V100 GPU
accelerated for AI/ML training
Machine Learning
AMIs
INFRASTRUCTURE
&
Greengrass
ML
Amazon Deep Learning AMIs
Compute intensive instances for
AI/ML Inference
C5
20. The AWS Machine Learning Stack
PLATFORMS
APPLICATION SERVICES
R E K O G N I T I O N R E K O G N I T I O N
V I D E O
P O L L Y T R A N S C R I B E T R A N S L A T E C O M P R E H E N D L E X
Amazon SageMaker Amazon Mechanical Turk
FRAMEWORKS KERAS
P3
NVIDIA Tesla V100 GPU
accelerated for AI/ML training
Machine Learning
AMIs
INFRASTRUCTURE
&
Greengrass
ML
Amazon Deep Learning AMIs
Compute intensive instances for
AI/ML Inference
C5
21. The AWS Machine Learning Stack
PLATFORMS
APPLICATION SERVICES
R E K O G N I T I O N R E K O G N I T I O N
V I D E O
P O L L Y T R A N S C R I B E T R A N S L A T E C O M P R E H E N D L E X
Amazon SageMaker Amazon Mechanical Turk
FRAMEWORKS KERAS
P3
NVIDIA Tesla V100 GPU
accelerated for AI/ML training
Machine Learning
AMIs
INFRASTRUCTURE
&
Greengrass
ML
Amazon Deep Learning AMIs
Compute intensive instances for
AI/ML Inference
C5
22. The AWS Machine Learning Stack
PLATFORMS
APPLICATION SERVICES
R E K O G N I T I O N R E K O G N I T I O N
V I D E O
P O L L Y T R A N S C R I B E T R A N S L A T E C O M P R E H E N D L E X
Amazon SageMaker Amazon Mechanical Turk
FRAMEWORKS KERAS
P3
NVIDIA Tesla V100 GPU
accelerated for AI/ML training
Machine Learning
AMIs
INFRASTRUCTURE
&
Greengrass
ML
Amazon Deep Learning AMIs
Compute intensive instances for
AI/ML Inference
C5
23. ML is still too complicated for everyday developers
Collect and prepare
training data
Choose and
optimize your ML
algorithm
Set up and manage
environments for
training
Train and tune model
(trial and error)
Deploy model
in production
Scale and manage
the production
environment
24. A managed service
that provides the quickest and easiest way for
data scientists and developers to get
ML models from idea to production
Amazon SageMaker
25. Amazon SageMaker
Collect and prepare
training data
Choose and
optimize your ML
algorithm
Set up and manage
environments for
training
Train and tune model
(trial and error)
Deploy model
in production
Scale and manage
the production
environment
Easily build, train, and deploy machine learning models
26. Amazon SageMaker
Pre-built
notebooks for
common
problems
K-Means Clustering
Principal Component Analysis
Neural Topic Modelling
Factorization Machines
Linear Learner – Regression
DeepAR Forecasting
XGBoost
Latent Dirichlet Allocation
Image Classification
Seq2Seq
Linear Learner – Classification
BlazingText
Random Cut Forest
ALGORITHMS
Apache MXNet
TensorFlow
Caffe2, CNTK,
PyTorch, Torch
FRAMEWORKS
S e t u p a n d m a n a g e
e n v i r o n m e n t s f o r
t r a i n i n g
T r a i n a n d t u n e
m o d e l ( t r i a l a n d
e r r o r )
D e p l o y m o d e l
i n p r o d u c t i o n
S c a l e a n d m a n a g e t h e
p r o d u c t i o n e n v i r o n m e n t
Built-in, high-
performance
algorithms
Build
31. • Training algorithm / inference code is
packaged in a Docker image
published on Amazon ECR
• SageMaker pulls the training
algorithm image from Amazon ECR
into the Model Training Service
• Amazon SM downloads or streams
the training data and runs the training
algorithm on the data.
• After training, Amazon SM uploads
model artifacts to Amazon S3
• For inference, Amazon SM pulls the
model artifacts and the inference
image from Amazon ECR, into the
Model Hosting Service
• Amazon SM exposes an inference
endpoint for client applications to send
prediction requests to check Fraud
• Ground truth data collected from the
client application could be sent into
the training bucket to retrain and
update the model
Deploying a Model on Amazon SageMaker