Philo-AD

Deep-learning based Anomaly Pattern Detection Solution

What is Anomaly?

Anomaly is defined as something that deviates from the normal behavior or what is expected. Anomaly detection is the identification of items, events which do not conform to an expected pattern Typically the anomalous items can be translated as an indication of intrusions, frauds, defects, error and etc. 

 

Anomaly Detection

 

AIM: Automatic detection of Anomaly in specific domain using AI Technology

Current Status: Automatic detection of Anomaly candidates in specific domain using AI Technology

 

Pain Points of Existing Technologies

 

1. Rule based approaches

- Need experts on specific domain to make rule sets

- NOT robust (Unable to detect slightly modified patterns)

2. Data-driven machine learning approaches

- Need vectorization of data records

- Users need to keep the entire data sets from its deployment for system upgrades

Philo-AD_eng.png

Philo-AD

Applies Deep learning artificial intelligence technology to effectively monitor suspicious behaviors and patterns from internal network as well as identify external threats. The process is in real-time and it supports instant update feature to maximize system reliability.

- Simplified AI deployment process with deep learning based time-series NORMAL DATA MODELING

   Enables a rapid deployment of the solution, saves at least 80% of the time for integration with 50% less cost

- Minimized alarm fatigues through Hybrid approach for anomaly detection 

  Deep Neural Network + HBKS (Augmented Pattern Memory)

- Simplified system update process by applying Model Management Framework

- Patented: Anomaly Pattern Detection System and Method 10-2019-0025066

Application

금융분야.png
Finance

FDS / AML

보안분야.png
Network Security

Network Intrusion Monitoring

제조분야.png
Manufacturing

Predictive maintenance

Smart Factory

Philo-AD Structure

1. Data Digitization

Convert non-digital data to digital vectors

 

2. Time-series Modeling

Applying a neural network model reflecting time-series characteristics

3. Hierarchical Behavioral Knowledge Space (HBKS)

Storing the anomaly patterns and its root cause identified by users, applying this information instantly for future analysis reference

Philo-AD_구조.png

Philo–AD Dashboard

 

Visualization of the data set

 

1. Monitoring & Graph

- Graphical presentation of monitoring, learning and verification process

- Analytic graphic presentation of anomalies

2. Auto Learning

Automatically collects data-set for automated learning

3. Explainable

Explaining the reason and the cause of suspicious patterns (Anomalies)