Applying Data Mining in Surveillance: Detecting Suspicious Activity on Social Networks

Introduction

Living in an atmosphere of peace and security is fundamental to human dignity and development. Security is imperative in every country, city and organization, but a state of security cannot be guaranteed. There are a million ways to bypass the system, but the risk can be minimized, and the opportunity given to offenders can be significantly reduced. Surveillance systems are used to deter, document, and are cost-effective ways to reduce crime rate. Even when you are away from your home, you can keep an eye on your property.

With the advancements in technology, the traditional surveillance systems driven by dedicated personnel, have now been replaced with automated and intelligent systems. These systems record data, have the ability to process it and capture information of interest to take instant decisions or remedial measures (Andrejevic & Gates, 2014; de Souza et al., 2016).

With these new capabilities in the surveillance systems, they have gained immense importance in the sectors of public safety and security. Numerous other industries such as physical security, health care, and science are also utilizing these systems for predictive analysis and behavioral analytics.

The systems should include continuous intelligence analysis or predictive crime analysis. It also includes video surveillance, cyber security, access control and fire detection. These kind of systems facilitate the surveillance of activities of suspicious groups and their networks by offering the ability to suppress data and manage data in a critical way. Body worn cameras, Record management and big data analysis are technologies that will drive a significant trend in this field. Processing huge data from surveillance system is a challenge to many organizations. However, there are several techniques to process big data. Data mining tasks can be dealing with such kind of challenges (Atrey et al., 2010; Bastani et al., 2016).

In this project, we focused on how to apply data mining tasks on surveillance system data and identify events of interest that are important e.g., object detection and tracking, detection of change and object classification human vs objects, analyzing suspicious networks and their activities and identifying nodes playing major rules in their activities (Bastani et al., 2016). The system level diagram of data mining-based surveillance system. It takes feed from different cameras to detect moving objects and then to classify them as human or non-human. It also performs surveillance of suspicious networks, monitors their activities and identifies nodes playing major role in group activities.

A centralized Surveillance system consisting of distributed components generates massive data from multiple sensors and locations. Key challenge is to handle this data and extract knowledge information that can be used by various industries for making actionable decisions.

With this challenge in sight, areas of focus are:

• •

  Automation for Data Handling

Data automation is necessary while tackling huge data to maintain the efficiency of the system. We need to work on automation techniques and methodologies for handling the captured data and processing it for further analysis.

• •

  Data mining on surveillance data

Useful information is extracted from raw surveillance data for the purpose of predictive analysis, behavioral analytics and decision making. Data mining becomes the prime area of interest in such systems as this becomes the source of information for the processing system.

Our paper aims to develop a methodology that would focus on extracting valuable information from surveillance data and using it for predictive analysis, behavioral analytics and decision making. This includes social network analysis image and detection of suspicious persons.

Related Work

The proposed research has two main modules. One is getting raw data from different sources and sensors and getting them processed to extract features and descriptors etc. The second and important phase of this research is data mining which is required to extract meaningful information from extracted features and making final decisions. This is important as the success of a project is mainly dependent on final decision. So, we have divided the literature review into two major parts. In the first part, the data mining algorithms for surveillance are discussed whereas the second part deals with the analysis of social networks.