The Sensor Cloud and Homeland Security

The Sensor Cloud

Today, cloud computing is no longer jargon, what with many organisations either evaluating or already running core enterprise applications in the cloud. Amazon’s EC2 (Elastic Compute Cloud), Google cloud services, Microsoft’s Windows Azure, and Oracle’s Public Cloud, are popular examples of cloud computing services; offering a variety of applications, platforms, and infrastructure for hire, from anywhere between an hourly basis to a fixed-duration contract basis.

The Sensor Cloud takes the concept of cloud computing and applies it to sensor networks: intelligent wired or wireless sensors store their data in the cloud, subscribers are allowed to view and analyse the data, and administrators carry out remote management of the sensors (if authorized by the owner of the sensor). The sensors may either be organization-owned, and therefore accessible only to authorized personnel in the organization; or they may individual-owned, with the data accessible to the public at large; or they may be government or organization-owned, with the data accessible to the public at large.

The concept of the sensor cloud has caught on because of the ubiquity of sensors consequent on:

• support for a variety of sensors on a standard smartphone platform (accelerometer, cameras, microphone, GPS, compass, proximity, ambient light, etc.)
• specialized sensors embedded in most electronic and electrical equipment: this feature coming about on account of the steep fall in the cost of such sensors

Ericsson predicts that there will be 50 Billion connected devices by 2020.

Companies are now building clouds to store the data captured by such sensors. MicroStrain recently announced that SensorCloud™, which previously supported only MicroStrain’s wireless sensors, would now support any web-connected third-party device, sensor, or sensor network through a simple OpenData API. In the US, Honeywell recently announced a new smart thermostat that connects to the Internet and uses Opower’s cloud service to record and analyse home heating and cooling demand.

Sensor Data Storage and Analysis on a CloudCredit: University of Washington, http://depts.washington.edu/dslab/SensorCloud/index.html

Homeland Security and the Sensor Cloud

How does these developments impact, or even matter to, Homeland Security planners in India?

There are several surveillance sensor networks in place or under implementation, in India; an overwhelming percentage of which are video surveillance networks, today. Each state government is responsible for setting up and running its surveillance sensor network; and different state agencies run their own networks (city surveillance, traffic management, VIP security, critical infrastructure, coastal surveillance, border control, etc.). In addition, other Homeland Security-impacting agencies have their own networks: Railways, airports, critical infrastructure, etc. Each network of sensors (video surveillance cameras in most cases) has its own infrastructure for video/content management (DVRs/NVRs/VMS Servers) and storage.

The problem with the current decentralized architecture is that many of the agencies do not have the required resources for managing the infrastructure; on account of which, older sensor data is not properly archived or is misplaced. In addition, silos of surveillance sensor networks prevent the emergence of a Common Operational Picture (COP) at an All-India level.

One solution that would address the above issues would be to build a Homeland Security India Sensor Cloud that services all Web-connected surveillance sensor networks: video, audio, radar, trace detectors, access control, motion detectors, etc. The Sensor Cloud will store the sensor data and allow authorized users to view and analyse the same. Agency-specific applications can be developed, by the Sensor Cloud administrators, based on the needs of the user-agencies. In order for such an initiative to get off the ground:

• States and their respective homeland security agencies have to be agreeable to having their data stored off-site (i.e. not under their physical control).
• Some sensor networks will need to be upgraded to meet a minimum quality-of-data level (for example, optical resolution in the case of video surveillance cameras). Such a move will also allow homogenization of sensor specifications, with the objective of ensuring that the most optimum sensors are selected for a particular use-case.
• The sensor networks will need to conform to a common standard (either by upgrading the hardware or by plugging in a gateway), to ensure that they connect to the cloud and that the data streams for each type of sensor can be stored in a common database. ONVIF has now become an IP-based physical security standards body, and it may be useful to specify conformance to these standards for any new sensor network.
• The owners of the data will have to be agreeable to sharing their data (in combination with data from other owners) for analysis, data-mining, and other information-extraction techniques.

Although creating such a Homeland Security India Sensor Cloud will be a complex undertaking, a start can be made by setting up a basic infrastructure and connecting existing Web-connected sensor networks on an experimental basis, with willing agencies. However, before doing so, the following issues will need to be addressed:

• Sensor Discovery: When there are a vast number of sensors on the network, finding a particular sensor within a specified time-frame is not an easy task. The Sensor Cloud’s data repository should allow easy discovery of a sensor, viewing the sensor’s data-stream, and administering the sensor.
• Security: When thousands of sensors start interacting with the cloud, with many of the sensors being intelligent devices with a native OS, the cloud needs to be made secure against threats of various kinds. The security system of the cloud should be in a position to verify that an identified sensor is actually what it claims to be, that it is carrying out its surveillance activity as prescribed, that it is recording actual events/parameters in its area of operation (rather than fudging data), that the data is being sent and has been received un-tampered, that no malware has been inserted in the data, that no unauthorized access to the cloud takes place, and that no unauthorized transmission of data and applications from the cloud takes place.
• Confidentiality: Since the data comes from various geographies and covers various physical parameters, the cloud has to ensure that sensor data, sensors, and associated information are not directly associated in the data repository. Furthermore, there needs to be a policy for the duration of time data is retained and for the destruction of data outside this retention period.
• Ownership: The ownership of the data may rest with the owner of the sensors or with the cloud operator or with both; while the ownership of information extracted from the data may rest only with the cloud operator. This has to be clarified up front.
• Applications and APIs: The process for developing specialized applications for different agencies subscribing to the cloud has to be laid down clearly. The cloud operator also needs to specify the APIs supported, that will allow sensors of different types, makes, and technologies to interact with the cloud.

Conclusion

This brief outlines a concept for a near-future-ready surveillance sensor network, that leverages current advances in cloud computing technology and sensor technology. The nitty-gritties of actually implementing the concept need to be thought through and researched before an experimental sensor cloud is created.

Homeland Security planners in India can make a couple of technological leaps, with respect to the homeland security infrastructure in the country, if they can work through the details and set up such a facility.