Published on 05, Jul 2023

Vijayasarathi Balasubramanian Speaks on How New Moves in Data Encryption Will Make AI More Secure

As more and more data breaches hit the news cycles, the conversation around privacy and protections kicked up into an all new high. Those who aren’t very tech savvy felt vindicated in their decisions to stay disconnected while others wondered what they could do to prevent things like this from happening in the future, and the public’s trust in larger corporations began to decline. This set an odd beginning as public opinion about AI and its capabilities began to sour. In a world of deep fakes and data breaches, day-to-day users began to worry about their overall safety online.

However, the one thing that’s commonly overlooked is how quickly and how hard data scientists, developers, and companies in the tech industry are working to close the divide. Vijayasarathi Balasubramanian, who also goes by “Vijay”, is one such expert who is keeping a close eye on data privacy. Vijay has almost twenty years of experience working as a data scientist for the retail and telecom domains, and so he understands how valuable data privacy can be, especially as more people The Next Billion come online and join the vast digital landscape.

“As more of data storage moves to the cloud and as more of our personal lives get integrated to online applications, data privacy stands out as the most urgent concern,” Vijay comments. “However, several companies are taking the initiative to improve this. Google Analytics 4 improves the way that users’ data is collected and used. Messaging apps like Signal, WhatsApp, and Telegram are investing in end-to-end encryption to protect users worldwide. The winds of change are here. We just need more companies to take this seriously.”

But what about how data is used when it comes to AI, the latest public spectacle? Rest assured that moves are being made to bring a higher level of data privacy to this space as well. AI is a very complicated system with multiple branches that can both interconnect and stand on their own.

Vijay himself is a Senior Data and Applied Scientist who uses AI, specifically its subset machine learning, to create recommendation systems for e-commerce. Recommendation systems are composed of interreliant algorithms that use large pools of big data to suggest products or services to users. Things have changed greatly since he first began to hone his skills, but it’s because he’s been so close to the change that he can understand the flaws within the system.

However, Vijay and several other experts agree that one of the best solutions currently available to ensure data security is homomorphic encryption, or HE.

HE allows AI to analyze and manipulate encrypted data without revealing what that data is. This allows for the preservation of sensitive data as it’s outsourced to cloud environments and increases the security of existing services. So, if there’s a data breach at a credit union, for example, HE would ensure that the data itself would remain secure even if the main system is compromised.

HE comes in several forms: partially homomorphic, somewhat homomorphic, and fully homomorphic encryption.

• Partially homomorphic encryption (PHE) is tractable, but it only allows specific mathematical functions to be performed on ciphertext. These functions can then be performed endlessly.
• Somewhat homomorphic encryption (SHE) only supports a select number of operations to be performed on ciphertext a set number of times.
• Fully homomorphic encryption (FHE) is the most complex and the most secure iteration of HE as it allows anyone to perform arbitrary computations on ciphertext any number of times and lets anyone without access to the security key so long as these operations can be efficiently computed.

Vijay comments, “These forms of HE matter because AI and its subsets intake high volumes of sensitive data from our face to our voice to our personal information. Recommender systems work off the machine learning framework and can track user activity and transaction histories to recommend products or services that appeal to that user. We live in a world where you can unlock your phone by showing your face. Because we are in such close proximity to our technology both physically and digitally, the rapid improvement and implementation of HE has become a chief priority.”

AI data consumption doubles by the year, and a growing number of people are gaining access to this technology. Because of this, it’s important to provide optimal protection to make sure that this data remains anonymous and encrypted. Though we are a long way from perfectly integrating data privacy practices uniformly, Vijay remains hopeful. More people are tuning into how businesses use their data, and more users are demanding control and accountability.

“Because of how quickly businesses have to adapt, those who take their time stand out in the cold and are heavily criticized for their unwillingness to change,” Vijay remarks. “I’m constantly looking for new ways to implement AI technology, and I’m always keeping a close ear on new trends in data science.  HE is creating new opportunities in the market, and though it has been in the making for the past ten years, its inclusion has a chance to radically improve AI, data science, and the Internet for years to come.”