Anomaly detection framework (ADF)

Detect anomalies in data

ADF (Anomaly detection framework) platform utilizes artificial intelligence algorithms and technologies to detect anomalies, suspicious actions, events, and unwanted behavior in large and complex data. The platform has a proven track-record across various domains.

top AI experts
30 %
preventing ~30% subscription frauds
years of delivering end-to-end implementations of AI systems

Keep decision-makers, operators and analysts in charge


Eliminate alert floods and alarms spam


Scale up the process by managing terabytes of data in parallel

Anomaly detection

Monitoring and analyzing of anomalous trends, patterns and events in processes is typically being done manually or based on a set of pre-defined rules. Such an approach leads to identifying mostly known patterns with large amount of irrelevant cases leaving hidden risks and unknown trends unspotted and unresolved.

Continuous improvement

ADF framework detects anomalous trends, patterns and events directly from all available data sources in real-time. That allows to uncover hidden anomalous trends or unknown patterns effectively and adjust continuously to anomalous behavior evolvement.

Detection time reduced from days to minutes or seconds.

Based on previous deployments, we can see that detection time to uncover both known and novel types of anomalies and threats is reduced to minutes or seconds.

How we reduce anomalies step by step

1) Anomaly detection

The system constructs a wide variety of time series and data clusters from small and large datasets. A rich algorithmic library then finds significant outliers, which are then analysed and reported.

2) Enterprise and external data

Various streams of data from different systems are used as an input source for AI algorithms to detect anomalies. Based on all available data we find new types of anomalous patterns.

3) Visualization

All detected anomalies are reported and presented in a user-friendly way to allow further analysis that focuses on relevant cases only.

4) Continuous improvement

The users can give their feedback, hence regulate the number of anomalies received as well as teach the system what is an anomaly and what is not.


Find inspiration for your business with our success stories



  • Find anomalous events and suspicious behavior.
  • Hidden risks management.
  • Concentrate on relevant cases only.
  • Detect new types of anomalous events in telecommunication data.


  • Novel cybersecurity threat of malware types detection.
  • Performance monitoring of critical devices.
  • Process streams of data from various systems.
  • Ensure smooth operation.
  • Reduce hidden security risks.


  • Predictive maintenance.
  • Production lines monitoring.

  • Reduce maintenance costs by detecting anomalous trends in production on early stages.
  • Increase efficiency in monitoring complex manufacturing and assembly systems.

Case study

How mobile service provider prevented 30 % frauds and secured ROI within 1 year

Despite implemented rule-based system and background check for new customers, our client suffered from subscription fraud (customers not paying for services and hardware) and was struggling to reduce it.

We have applied our own machine ;learning fraud detection framework to enhance risk management and empowered fraud prevention process with AI.

We have completed comprehensive journey from idea and proof-of-concept to deployment of ML fraud prevention process, preventing roughly 30 % frauds, which secured ROI within 1st year.

Area: Telco, credit risk


Do not miss

If you want to reduce detection time from days to minutes or seconds, contact us:

Thank you

We will contact you as soon as possible.

Ondřej Vaněk

CEO Blindspot Solutions

Štěpán Kopřiva

CTO Blindspot Solutions