Anomaly Detection & Risk Management
Detecting anomalies is a crucial capability across nearly all industrial sectors.
In the field of information security, for example, this involves automatically identifying unusual user
behavior from target data to detect potential threats preemptively. Security models are tailored to suit
various working conditions by analyzing real-world data. The manufacturing industry analyzes structured
and unstructured data from multiple equipment sensors to develop optimal anomaly prediction models
to achieve predictive maintenance and minimize downtime.
Let us take a closer look at the functionalities applicable to the security industry:
① Automatic Risk Analysis in Emails and Chats
– Filters messages containing specific keywords or patterns to identify risk factors
– Analyzes the tone and severity of message content to categorize risk levels
automatically
② Key Information Masking
– Extracts and masks data containing personal or sensitive company information from
databases
– Automatically recognizes and masks entities such as names and addresses based on
context
When using public LLMs, to prevent data leakage, personal or sensitive company information is
converted
into dummy data before being processed by the LLM and then restored to the original
format upon receiving results.
③ Unstructured Data Security: Images and Attachments Management
– Extracts only the paths and embedding vectors of attachments for storage and
management in the database
– Identifies and analyzes text and tables within images or documents to assess risk
④ Monitoring Anomalies Through Transmission Pattern Analysis
– Analyzes patterns based on historical data to detect anomalies
– Automatically identifies and flags messages with atypical tone or content, generating
alerts
In addition to the built-in scenarios,
customized scenarios according to customer requirements can be applied within a few weeks.
Inquiry
EN 
KR