Cyber Security and Streaming Data: IoT, AI and Security Policies

Many companies have cybersecurity policies that inhibit their ability to send data outside of company networks. These days, companies are struggling to maintain operations with the current constraints of remote work, reduced staff, or both. Monitoring systems remotely has become even more critical. However, current systems and corporate policies make this very challenging for many industrial companies. Here I explain how companies can get started with advanced analytics given current policies, until such policies may evolve to enable greater flexibility – without giving up security.

CHALLENGES WITH CYBERSECURITY

Cybersecurity is necessary to prevent tampering with data that could lead to production loss, equipment damage, product manipulation or loss of intellectual property and sensitive data. However, companies must balance between tight cybersecurity policies and the latest technological advances, which are vital for productivity gains. There are several inhibiting factors that drive companies to clamp down with the strictest possible security policies:

• They lack the latest technical understanding: Many companies haven’t yet experimented – let alone embraced – the latest analytics technologies and don’t have the full understanding of the related complexity and potential. Due to this lack of certainty, inertia is inevitable: companies delay onboarding new technology and keep data within their network to feel comfortable and secure.
• They lack capacity: In many cases, the company doesn’t have the bandwidth to tackle new projects, such as advanced analytics implementation.
• Out of date policies: In some cases, the technology is changing faster than the company’s policies. Policies might not be updated on a regular basis making it difficult to keep up with development. Moreover, changing corporate policies requires significant effort.

SOLUTIONS

In order to tackle this, you have two options:

1. Convince your company to change policies.
2. Accept the policies and try and find solutions within the policies.

Typically, companies are most interested in accessing continuous operational data in order to feed a model. Historically, this model would be a physics-based simulation, but increasingly could be a data-driven statistical model, or a hybrid of the two. Some models need to be trained and thankfully most such models can be trained offline. Corporate security policies on streaming data generally can be avoided in this initial model configuration. The solution for deploying the trained model is edge analytics. In a previous article, we defined edge analytics as follows:

""Edge analytics is an approach where you collect and analyze data at a non-central point, such as the sensor, device or touchpoint itself. Analyzing data as it’s generated lowers companies’ costs and increases the ability to drive timely, valuable business insights from all of the data to inform actions and decisions at remote sites.""

This approach can avoid the challenges of continuously streaming data outside of a plant or operational setting.

FINDING SOLUTIONS WITHIN YOUR POLICIES

You have two options for finding solutions within your policies:

1. Running models on edge agents: Let’s say you have ten different plants or vessels from which you collect data. If you can transfer a batch of this historical data offline or securely online within your data policies, you can use that data to train your models. A trained model can then be deployed locally on an edge agent at the operational site. Each plant will get an edge agent using the same model you have trained, but the operational data will never leave the plant site.
2. Running the model on the edge gateway: If you have a private company network, the edge agents for these 10 plants or vessels can stream data to an edge gateway – a computer connected inside the private network. In this mode, a central model will run locally on the edge gateway with data from all ten edge agents. This way, the data doesn’t have to leave the network, but instead of ten local models, there can be one global model.

FOCUS ON THE BUSINESS PROBLEM

We cannot stress this enough, but before you decide how and where to deploy edge models, focus on the business problem you need solved. Identify what are the highest value and feasible use cases. If you cannot change your policy, by bringing analytics to the edge inside your existing network, your company can still start utilizing your data to influence decisions and generate greater value.