How an Ex-NSA mathematician changed the way Covenant Eyes monitors porn use

The online pornography accountability service Covenant Eyes released this year a new software principally designed by a former National Security Agency data scientist that makes it nearly impossible for users to take advantage of loopholes in order to view pornography undetected. 

As hundreds of thousands of people have used Covenant Eyes to hold themselves accountable since the company’s founding in 2000, the organization switched its entire offering over to its new “Screen Accountability” software in March.

The new software is billed as a “revolutionary screen monitoring technology” that is a “radical new approach to accountability” and vastly more effective than the company’s old system, which relied on text-based detection through web browsers. 

As the way people use the internet is constantly changing, the company's new detection software uses an image-based detection system that can detect pornographic images and videos regardless of their source, online or offline: an encrypted web browser, a mobile app, an external flash drive or even a Microsoft Word document. 

“The software sees the screen just as you see it,” Covenant Eyes Chief Data Scientist Michael Holm told The Christian Post in a recent interview. “The battleground has moved from the heavily-hampered world of network text-parsing right up to the visual input to your eye." 

Holm started working for Covenant Eyes in 2014 after serving three years as an applied research mathematician for the NSA, a federal agency headquartered in Maryland.

Holm began working for the NSA in 2011 after earning a doctorate in mathematics from the University of Nebraska. He said he desired to be part of a mission bigger than himself and to get a start in the competitive industry. He found his work at the NSA to be “purposeful and valuable."

However, desiring to move his family back to the Midwest where most of his extended family lives, Holm began investigating new job opportunities about three years into his employment with the federal intelligence agency. 

But the very weekend that Holm intended to accept the job offer, the Walmart recruiter happened to be away at a conference and during this time, Holm was informed by his wife about a job opening at the Michigan-based Covenant Eyes.

Holm, who grew up in a Christian household, was loosely familiar with Covenant Eyes at the time.

“During the interview, I was greatly impressed with both the mission and the quality of the company, and it seemed clear beyond any doubt that this was where I belonged,” Holm explained. “It was a great fit in terms of using my gifts in a mission close to God’s heart. So, I accepted the job fairly-well on the spot and ended up turning Walmart down.”

Holm was initially hired by Covenant Eyes as a data scientist at the end of 2014. In his first six months, he spent time investigating the possibility of using an image-based system to detect pornography and eventually came to the conclusion that it could be done. 

“With the fast growth of encryption in the digital marketplace, the increasing use of apps over browsers and the growing diversity of operating systems and platforms, the text-based system was quickly becoming obsolete,” Holm explained. “And, therefore, the company would not have been able to continue very long that way.”

“Image-based pornography detection was a huge conceptual change for Covenant Eyes at the time. And while I didn't yet know it, God had put me in that place at that time for a purpose higher than myself, just as I and others had desired and prayed for, but equally outside of my personal design or anticipation," he continued. “And I can’t take credit for God’s timing or what He was accomplishing during that time – in mid-2015 the technology was ripe and just beginning to be capable of the tasks that we needed it to perform.”

For two years, Holm said, he worked alone on the project developing and optimizing mathematical machine-learning models capable of distinguishing pornographic images from nonpornographic images.

“In a sense, an image is raw data. More precisely, one may think of an image as a highly-structured, three-dimensional ‘blob’ of numbers between zero and 255,” Holm explained. “And so, my job was to develop an algorithm to accurately and consistently distinguish those three-dimensional number blobs depicting pornography from that three-dimensional number-blobs of numbers depicting [non-pornography]. At its core, this is a mathematical optimization problem in a very high dimensional feature space, and today, this task can be done well." 

As the capability matured and showed a track record of success, Covenant Eyes began to build around the project. 

With “Screen Accountability” now fully implemented, Holm warns users trying to skirt the system that there is “very little avenue to get around the new approach compared to what we had before.”

“I think that one of the most significant improvements in this whole switch is that Covenant Eyes users who used to have tricks up their sleeves to get around the software, they will no longer work,” he stressed. “It's a totally different ballgame. And very few, if any loopholes.”

On the basis of the artificial intelligence work that Holm has done for Covenant Eyes, a daughter business affiliated with Covenant Eyes has been launched. 

The name of the business is Picnix with a mission to provide similar AI-driven solutions in other industries. 

For example, over the last two years, Picnix has developed a state-of-the-art machine-learning-based system to detect new oil and gas reservoirs using regional 3D-seismic data. To date, the company has successfully identified oil and gas in Michigan and Texas.

“We have a small but growing team of data scientists now but we continue to support the original work at Covenant Eyes, even while developing these other new areas. We're sister companies,” he said, “and I am really part of both companies right now.”

Finding success in his career and enjoying his life in the Midwest, Holm pursued a childhood dream in Michigan: taking part in his family's farming heritage. 

"I've had farming in my family for several generations and now have the opportunity to take part in that heritage through raising my family with a variety of endeavors like chickens for eggs and cows for milk," he said. "This, however, together with my work at Covenant Eyes and Picnix, has been a significant growing and stretching
experience for me, and I am very often bumping into the awareness of my limits. So, I am most thankful to God for His grace and higher purpose in all things, whether in me, through me or in spite of me.”

Holm and his wife, Hannah, who met in college, are raising their six children, three of whom were born in Michigan. 

Comments are closed.