By Jumbi Edulbehram

Open up any tech publication and you’ll find new and innovative ways that artificial intelligence (AI) and machine learning are permeating every aspect of our lives. This week alone, one company announced that it was battling fraud using AI and another is analyzing Twitter feeds to find romantic matches. So why are AI and machine learning getting so much buzz in security?

I’ve been involved in the analytics world since 1998 when arguably some of the traditional applications were being introduced. What we’ve discovered over the last 20 years is that traditional analytics have a hard time interpreting real life, but an easy time following instructions.

For example, a basic application like a tripwire works by implementing an algorithm to alert users of when anything crosses the line. The issue becomes whether that alert is set off by a human threat or other more random things, like an animal or the wind. These kinds of mistakes can result in missed alerts.

Incorporating AI and machine learning into these applications allows the computers to actually understand the real world in a more robust way. In fact, unlike in the ?old days? when you had to tell the computer what to look for, in the new world of AI, you can give a computer a whole bunch of examples of what a person is doing in many different ways —whether they’re walking, crawling, running, flying, etc.— and it can decide for itself whether it’s a person or not.

Even in the last few years, people have realized this is a much more reliable way of interpreting real threats, which is why AI is starting to generate a lot of buzz in the security industry. This is fundamentally shifting how video data is analyzed and acted on.

What we’ve discovered over the last 20 years is that traditional analytics have a hard time interpreting real life, but an easy time following instructions.

There are several applications that are set to benefit from AI and machine learning within the security industry, as well as a range of companies implementing interesting applications and ways of using the information:

• In more traditional security applications where users are applying machine learning to make alerts more robust and accurate. For example, adding an additional layer of intelligence to the previous tripwire example helps improve the accuracy of the alerts being generated.
• Analytics such as facial and license plate recognition in traditional security. With new technology implemented, this software becomes much more accurate and extensible to not only recognizing particular faces and identities, but also classifying identifiable characteristics, such as age, gender, ethnicity, etc. This is much more feasible to do when you can teach the computer various methods by which to apply these categories. In addition, this opens up a host of other applications, including biometric applications, like image and voice, which is useful in access control.
• Anomaly detection or incident predictions, which didn’t exist in the previous world of traditional analytics; this is where AI and machine learning really will have an impact on security. There are companies now, for example, that have developed software that can analyze the pixels in every image of a video and correlate that data with changes in every other pixel. Once the computer looks at that dataset for a while, it can identify anytime an anomaly occurs.

Where this technology becomes really interesting is in how its applied through the use of robotics and drones, which can be utilized in situations where they’re trusted to make decisions based on the information being collected and humans aren’t needed. In the ?old? traditional ways of using robotics, many of the applications were in diffusing bombs or gathering intelligence, and now, AI and machine learning allow this technology to enter situations and make certain decisions based on the information being collected. The implications of this can be virtually endless.

In essence, the security industry shouldn’t be afraid of using AI and machine learning to its advantage, as the technology advances and adds significant value for end users. The challenge now is in overcoming the stereotypes associated with the use of these applications and shifting the thinking toward more long-term and focused cost-benefit-analysis for implementing such tech trends.

About The Author
Jumbi Edulbehram is a thought leader within the security and business intelligence markets, and serves as the Regional President ? Americas for Oncam. His career has spanned more than 30 years with senior leadership roles at Intel, Samsung Techwin, Intellivid, and Axis Communications.