Abstract
Cyber Patrol is software to filter the internet and block content based on one or more categories: nudity, alcohol, etc.
Software of this type is used not for people to protect themselves, but for authority figures to control what others may view. Cyber Patrol targets many markets: businesses, schools, proxy servers, and even governments. (Cyber Patrol partners with Prodigy to provide censored internet access to China.) Because of this, it has earned the epithet "Censorware."
And because nearly all existing products refuse to reveal the contents of their lists, calling them trade secrets, those lists have earned the epithet "Blacklists."
With
a hundred million pages already on the web, and thousands of new
domains being added every day, how can any one group hope to find
all the material which should be blocked, or even most of it? The
human resources required just to examine new sites on the web are
staggering, not to mention the additional task of reexamining existing
material.
In the race to advertise a larger and larger database, it is easy to forget that errors made in haste may stay on the secret list for months, or years, or may never be found. There are over 50,000 entries in the CyberNOT database, and CyberPatrol does not notify the sites it blocks.
This report examines some of the thousands of sites which Cyber Patrol blocks in their entirety, against two criteria: first, that blocking should be accurate (e.g. a site should not be blocked as sexually explicit if it has no such content), and second, that blocking should not be overbroad (e.g. because site A has explicit material, site B should not be blocked based on a similarity of name or sharing a computer).
Our conclusion is that Cyber Patrol blocks a great many sites which do not deserve to be, and that furthermore, looking at past reports of the product's accuracy, fixing these errors is a low priority.
The pattern is the same: they are shocked, even outraged. They defend themselves, saying they can't imagine why anyone would think such a thing.
"...all it would have taken was a few minutes of investigation on the part of Microsystems to find out about the Neighorhood Watch program at GeoCities..."
"more than 50 sites are hosted on the machine; this means more than 1000 interesting pages won't be available to anyone using this piece of crap software."
misc.headlines SexActs Violence/Profanity
misc.health.injuries.rsi.moderated SexActs Violence/Profanity
soc.feminism SexActs Violence/Profanity
talk.abortion SexActs
rec.games.chess.analysis Quest/Illegal/Gamble Violence/Profanity Intol
Might Cyber Patrol 5.0, or 6.0, be a spectacularly better product that will eliminate all our concerns?
CyberPatrol's wildly overreaching blocks ... clearly establish that the Cyberpatrol approach to censorship is not "narrowly tailored."
"Why does it matter what is blacklisted? The librarian could just override the censorware for that particular site."
As hard as companies like Microsystems may try, filtering the Internet is a challenge similar to filtering the ocean.
If Cyber Patrol can't deal with 67 specific blocks handed to them on a platter and with public scrutiny, how well can they maintain a database with (currently) fifty-eight thousand secret, unreviewed entries?
In this report, we do not compare Cyber Patrol against other products of its type. In fact, it may be one of the better censorware programs (it has certainly won enough awards). We take no position on its value relative to competing products.
Reporters: note that the list of blocked sites in this report has been verified by independent, unaffiliated journalists in advance of publication. Please contact the authors for more information. The press release associated with these webpages is also available.
Ada and Yoyo are two sites blocked by Cyber Patrol. Details are in the Introduction.
Update: as we predicted, most of the blocks we describe in this report were quickly unblocked by The Learning Company. We were surprised to learn that eight or ten of the sites we mentioned were not unblocked, though they clearly should have been, and that the hundreds of newsgroups were not even reviewed. Our aftermath update is now available. 12/25/97
Several members of the volunteer organization Peacefire were instrumental in helping to prepare our data. We owe thanks to Nicole Carlson, Reid Conti, Lee Dia, Benjamin Jenkins, Seth Schoen, Ross Sieber, and our other (anonymous) assistants.