Keystroke Behavioural Analysis For Fraud Detection: Deep Learning as-a-service Infrastructure

Transcript

1. Keystroke Behavioural Analysis For Fraud Detection: Deep Learning as-a-service Infrastructure arbitrio@fbk.eu __pamaron__ Ernesto Arbitrio
2. Who am I
3. The problem User Identification Biometric Identifiers
4. The solution: data science perspective Keystroke dynamics: analysis how a user types by monitoring keyboard inputs thousand of times per second, and processing this data through an algorithm, which then defines a pattern for future comparison Keystroke Dynamics Identifying an individual based on their way of typing on a physical or virtual keyboard
5. Keystroke Dynamic Analysis
6. Individual DeepKS Learner
9. The solution: software eng perspective customer A bucket of RESTful APIs directly connected with the extraction feature module and the prediction model API Engine Feature extractor DL Model {json} Raw data, features, predictions
10. API Engine: The TOOLS Eve http:/ /python-eve.org/ Luigi https:/ /github.com/spotify/luigi REST API framework powered by Flask and Cerberus and it offers native support for MongoDB data stores. Luigi is a Python module that helps you build complex pipelines of batch jobs. It handles dependency resolution, workflow management, visualisation etc.
11. Why a specific tool for the pipeline? - Task templating - Dependency graphs - Central Scheduler - Resumption of data flows after intermediate failure - Command line integration - Error emails
12. Luigi Concepts Dependencies Targets Tasks Parameters
13. The target
14. The task
15. The central scheduler