You are here

‘Big-data’ crunching needed

Category: 

OTTAWA—Federal security agencies risk being overwhelmed by threats—or failing to even foresee them—unless they embrace the digital-age phenomenon of big-data crunching, warns an internal Public Safety Canada presentation.

With billions of people using mobile phones and surfing the Internet, security officials are preoccupied with getting timely access to valuable information.

But the recent past is littered with challenges related to information sharing and privacy—including controversy over the Conservative omnibus bill known as C-51, the presentation notes.

Officials acknowledge the public might not trust government to respect privacy in the process, pointing to revelations by former U.S. spy contractor Edward Snowden about widespread surveillance of communications.

“Canadians are increasingly concerned about issues of crime and terrorism, but in the post-Snowden era, public concerns about government data use may stand as a barrier to the effective use of innovative data analytics by law enforcement and security organizations,” the presentation says.

Big-data analytics generally refers to the process of gathering and systematically sifting through millions or even billions of pieces of data—numbers, text, graphics, videos, and sensor information—to glean insights that can’t be detected through standard methods.

Given the pace of technological advancement, and the exponential increase in the amount of data produced worldwide, there may be opportunities to do things more efficiently and identify patterns through innovative techniques, the presentation says.

“In other words, once we have the information, how can we ensure that we are in a position to use it effectively?”

The May presentation, “Big Data Analytics in the World of Safety and Security,” was prepared for Public Safety Canada’s internal policy committee.

The Canadian Press obtained a declassified version of the secret draft presentation under the Access to Information Act.

Small portions were withheld due to their sensitivity.

The presenters cite examples of using big-data analysis to create efficiencies, find a needle in a haystack, and fill data gaps.

For example:

  • Philadelphia police mined data to predict a parolee’s risk of re-offending to determine the necessary level of supervision;
  • U.S. researchers found a genetic variant related to schizophrenia was not detectable when reviewing 3,500 cases but were able to pinpoint a trend by looking at 35,000 cases; and
  • In Guatemala, a pilot project revealed how mobile phone movement patterns could be used to predict socio-economic status.

More than two years ago, Jennifer Stoddart, the federal privacy commissioner at the time, cautioned that big data had not simply increased the risk to privacy, it had changed the very nature of that risk.

The Public Safety presentation allows that privacy considerations and building public confidence must be taken into account.

“Privacy does not need to be a barrier to innovative data analytics,” it says.

“We need to think strategically about what we want to accomplish with data and then design in appropriate privacy protections.”

The officials also stress the importance of finding the right partners to pursue “promising practices,” as well as ensuring agencies have the technology, policies, and people to make the most of the techniques.

Twitter icon
Facebook icon
Google icon
LinkedIn icon
Pinterest icon
Reddit icon
e-mail icon