Headquarters

 

DomaniSystems Incorporated

2 Enterprise Drive, Suite 301

Shelton, CT 06484 USA

 

info@domanisystems.com

Phone:  1-203-567-0099
Fax:  1-203-567-0624

Search
  • Dale Montrone

Blockchain Meets Machine Learning and IoT - Power of Three

Future is coming at us very fast! This future will be based on convergence of three technologies: a) Blockchain, b) IoT , and c) Machine learning [1].


The world surrounding us consists of huge IoT (Sensor) networks recording every action we take. AI technology can help in analyzing these actions, detecting anomalies, identifying security threats and in taking appropriate measures.


Blockchain technology when combined with this vision of AI powered IoT networks has the potential to provide a secure environment with an immutable audit trail of events.


In this blog, we are going to expound on the use of AI in anomaly detection. In subsequent blogs, we will cover structuring and optimizing IoT networks.


So, What Is Anomaly Detection?


Anomaly detection is a technique used to identify unusual patterns that do not conform to expected behavior, called outliers [2]. It has many applications in business, from intrusion detection (identifying strange patterns in network traffic that could signal a hack) to system health monitoring, and from fraud detection in credit card transactions to fault detection in operating environments.


A good survey on anomaly detection technique is in [3]. The most common techniques fall into the scope of statistics, clustering and machine learning. Depending on the types of samples necessary to process the data, these techniques are divided into supervised, semi-supervised or unsupervised.


Supervised techniques require a training dataset with labels indicating the category of each sample (e.g., ‘no attack’, ‘jamming’ or ‘selective forwarding’) [4]. Then, a model is generated to classify new unlabeled samples into one of the defined categories. Semi-supervised techniques require a training dataset with samples of a single category in order to create a model that classifies new samples as belonging to that category or not. Finally, unsupervised techniques do not require labeled training data and are capable of dividing a dataset into various subsets without a previously learnt model.


Depending on the characteristics of the specific scenario and on the requirements of the application, some algorithms perform better than others [5].


DomaniSystems’ security approach for IoT networks combine Machine Learning with Blockchain technology.


References:

[1] https://www.forbes.com/sites/oracle/2018/01/10/iot-ai-and-blockchain-time-to-reimagine-the-art-of-the-possible/#5648f5253107

[2] https://dzone.com/articles/anomaly-detection-in-mobile-sensor-data-using-ml

[3] Victor Garcia-Font , Carles Garrigues and Helena Rifà-Pous. A Comparative Study of Anomaly Detection Techniques for Smart City Wireless Sensor Networks., June, 2016 (file:///C:/Users/Dae%20Han/Downloads/sensors-16-00868.pdf)

[4] Kim, I.; Oh, D.; Yoon, M.K.; Yi, K.; Ro, W.W. A Distributed Signature Detection Method for Detecting Intrusions in Sensor Systems. Sensors 2013, 13, 3998–4016

[5] Chandola, V.; Banerjee, A.; Kumar, V. Anomaly Detection: A Survey. ACM Comput. Surv. 2009, 41, 15:1–15:58.

29 views