In the previous article I discussed how the distributed computing technology called Hadoop marks a big difference between the Amazon’s AWS and Facebook’s respective types of cloud computing.
In this article I want to connect Facebook’s success at scaling big data processing for the social web with the earlier discussion of face recognition technology.
Your Face Here
For a face recognition system to work it must have access to a database of identified images against which to process new captures for possible matches. So, for example, that scene in Minority Report where Tom Cruise’s character receives personalized ads as he walks through a commercial space implies a database that includes at least one image that is already associated with the name ‘John Anderton’. And we can reasonably surmise that the database, like a utility power grid, provides pervasive access to authorized software—advertising, law enforcement—that might want to match images streaming in from camera sensors in the movie’s ubiquitously-computing future society.
In our time, US Senator Al Franken reminds us in testimony from a Judiciary Committee hearing in 2012 that “Facebook may have created the world's largest privately held database of face prints without the explicit knowledge of its users." And that lack of explicit user consent is why EU countries (Norway, Germany) have forced Facebook to disable their “photo tag suggest” feature, which uses face recognition analysis to link faces with names of other users in the Facebook system. With the feature turned-on, as it is by default, and when the face recognition software finds matches, the relevant users are instantly tagged in photos their Facebook friends upload; all it takes is a click for the uploading user to accept the suggested tags, kicking off a notification to all those tagged.
Facebook stipulates in its latest user agreement that it can use photos in its system for advertising purposes. We should assume that “advertising purposes” includes granting access to third parties that very much would like to advertise to its potential customers based on being able to review and analyze images of them in one or more scenes from everyday life.
If this flow remained entirely benign, all you might notice of this process is that ads you see online are indeed more personally appealing and relevant to your interests.
Access to the world’s largest database of already-tagged images is a law enforcement dream. Increasingly, Federal officers make that dream come true with a warrant. In a twist, rather than seeking to see images from the Facebook database, some courts have used the popularity of Facebook as a form of public shaming, setting up an account and posting DMV pictures of people with outstanding warrants in their jurisdiction.
Here technology, law, law enforcement, commerce, and politics begin to crash into each other. If the photos are nicely tagged and—thanks to Hadoop--easily findable, then posting a photo to Facebook means releasing it not only to anyone in your circle of “friends” but to anyone who has, can pay for, or can get warranted access.
If you encrypt an image and upload it to Facebook then the system can just copy it when your friend decrypts it for viewing. Since Facebook has a vested interest in having access to user images for those “advertising purposes,” we should assume that the system copies any uploaded images when “friends” decrypt them. And if so, Facebook image data is always in the clear when it becomes part of the system.
Encryption for Monitoring Tools
Fortunately, as we see fit, we can encrypt image data within our own networks and trust they will remain that way except at the intended endpoints. And we can ensure that our monitoring systems respect our data encryption policies. For example, SolarWinds Network Performance Monitor can poll for MIB updates on network devices through SNMPv3, keeping data on the state of your network confidential as it passes through its packet-switched route.