Cloud, AI, And Change Of Mindset: The Security Of Today And Tomorrow

Cloud, AI, And Change Of Mindset: The Security Of Today And Tomorrow

Among the technologies considered at the same time mature and suitable with respect to the needs of the security sector, two stand out: video surveillance as a cloud service (a phenomenon that started 10 years ago) and artificial intelligence (to classify objects, for forensic purposes, to reduce false alarms or for business intelligence solutions).

Cloud Video Surveillance

Among the main advantages from a technological point of view is centralized control, the fact of being able to update multiple workstations from a central location (without sending the operator on site), and remote control. However, much-improved functionality due to the covid, which has pushed digitization. But the cloud had a major impact on the business model, as it shifted the buying process from capital expenditures to operational expenses. This offers end users greater flexibility with respect to the purchase of technology, which is actually rented/borrowed according to the needs of the moment. This generates different revenue streams than in the past, being monthly but continuous fees: the vendor will therefore have a regular cash flow and will be able to upsell the products by offering different additional services. The same integrators will have to change their way of doing business and, above all, their mindset by entering into a management logic rather than a sales one. 

Artificial intelligence

But the real star of the investigation is obviously artificial intelligence. If the automatic video analysis made it possible to detect any anomalies (unattended bag, improperly parked car, crossing the line) and then alert the operator, the AI ​​solves the problem of false alarms, thus automating security processes on a large scale. But there are also some gray areas to evaluate before implementation. 

Where is the data located?

Let’s start with the cloud, where the greatest risk is data localization. How can we be sure that the cloud provider is providing us with a secure service that prevents user names and access rights from being shared externally? And if a third party checks my data, how can I mitigate contract change costs? What will prevent a provider from charging me thousands of dollars to switch to another provider and get back the data he has stored? Or, coming to a very current example, with the conflict in Ukraine, it could happen that a cloud provider suddenly cut off all the cameras produced in a certain part of the world: you would find yourself with systems that do not work out of control. All issues are to be taken into consideration.

New questions to ask

Artificial intelligence involves a total mindset shift for security system integrators. It’s not just a matter of understanding the hardware but also of understanding how the underlying software works. System integrators must ask at least five questions to their suppliers:

1. The first set of questions aims to verify the AI ​​performance as claimed by the manufacturer. Does the performance meet customer requirements? Is it easy to fool? What happens if you wear a disguise? Will the false alarms return?

2. The second set of questions is about privacy. How does AI fit into the privacy and data protection measures? The more advanced the AI, the more metadata is collected. Age, gender, what you wear, the car you drive, make, model and color are all collected and stored. Does all this expose you and your customers to any regulatory compliance risk?

3. Can you explain how AI works? Is this a magic box, or is it possible to explain why the algorithm made a specific decision?

4. The fourth set of questions concerns possible bias: Is there any bias in the way AI controls the outcome? For example, if the algorithm was only trained on Caucasians, what would it do when dealing with Asian facial features or people of color? Will the result be distorted? What is the quality of the dataset used to train the AI? 

5. Camera manufacturers offering AI rarely have an in-house team that develops algorithms. They typically use the training sets and algorithms available on the market. Therefore, when evaluating a new AI system, you need to ask what kind of dataset it uses, what its origin is, and whether it was lawfully obtained.

The Clearview AI case speaks volumes: can using such a solution get me in trouble? Where does the training set come from? Where does the algorithm come from? Could two vendors use the same training set and algorithm? And if so, how do they differ?

In the market of the future

In the coming years, software development capabilities will be the real focus: competition between vendors will be based on software capabilities rather than catalog size or price. Many AI companies will enter the market, and the competition between them will revolve around three main elements.

1. Metadata. Which vendor can give us better metadata extraction from video feeds or access control systems? How many attributes can be extracted from the image? What level of detail can be achieved?

2. Quality of inference. As you know, the same things can appear very different with varying light conditions. For example, a silver car can look white at night. Algorithms that can guarantee more accurate results and provide better inference from their engine will benefit.

3. Discover and create links between different attributes. One of the key uses of artificial intelligence today is in forensic research. Now we can type in the system: “I’m looking for a man with a blue shirt and black pants.” The next step will allow the system to automatically identify which car it arrived in, the vehicle’s make and model, and the license plate number. 

Artificial Intuition

Finally, we come to the Holy Grail of artificial intelligence solutions: artificial intuition. The human brain is able to make decisions even in totally new situations, thanks to experience and instinct. For algorithms, this is not possible. Not yet, at least, because as technology evolves and neural networks are refined, we will be able to see computerized systems with a certain intuition or the ability to understand new situations and therefore decide, independently, the best course of action. And this will naturally also open up an ethical issue.

Also Read : What bullish Candlestick Patterns Are And How To Use Them To Buy Stocks

stuff In Post Team

Stuff In Post is one of the top tech news and updates websites. Our platform is a hub that provides all the trendy and accurate information on time. We also publish the latest updates on Business, Marketing, Finance, Gadgets, Software, and Apps, along with Technology.

Leave a Reply

Your email address will not be published. Required fields are marked *