Published January 13, 2015
Cox Communications appears to be interfering with file-sharing by its Internet subscribers in the same manner that has landed Comcast Corp. in hot water with regulators, according to research obtained by The Associated Press.
A study based on the participation of 8,175 Internet users around the world found conclusive signs of blocked file-sharing connections only at three Internet service providers: Comcast and Cox in the U.S. and StarHub in Singapore.
Of the 788 Comcast subscribers who participated in the study, 491, or 62 percent, had their connections blocked.
At Cox, 82 out of 151 subscribers, or 54 percent, were blocked, according to Krishna Gummadi at the Max Planck Institute for Software Systems in Saarbruecken, Germany.
Philadelphia-based Comcast is the country's second-largest ISP, with 14.1 million subscribers. Atlanta-based Cox Communications is the fourth-largest, with 3.8 million. It is part of privately held Cox Enterprises Inc.
Comcast's practice of interfering with traffic was brought to light by user reports last year and confirmed by an AP investigation in October.
Consumer advocate groups and legal scholars criticized the interference, saying that letting an ISP selectively block some connections makes it a gatekeeper to the Internet. Their complaints prompted the Federal Communications Commission to launch an investigation, which is ongoing.
Legislation also has been introduced in Congress to guarantee "Net Neutrality," or equal treatment of traffic by Internet service providers.
Comcast maintained that the intervention was necessary to ensure that non-file-sharing traffic would not be impeded by a few heavy users of file-sharing programs like BitTorrent. But in February, it said it would stop selectively targeting file-sharing later this year.
Much of the FCC's attention to the matter has been focused on Comcast's secrecy — before the AP's investigation, it acknowledged only in the most general terms that it was managing traffic.
At least since 2006, Cox's subscriber agreement has noted that the company engages in "protocol filtering," which means that it treats different types of Internet traffic, like Web surfing, e-mail and file-sharing, differently.
"To ensure the best possible online experience for our customers, Cox actively manages network traffic through a variety of methods including traffic prioritization and protocol filtering," the company said it a written statement.
Cox denied that protocol filtering amounts to discrimination of any specific services.
The blocking observed by Gummadi's group occurs when a subscriber has downloaded a file using the BitTorrent application and tries to upload it, or share it with others, over the Internet connection. The main victims are the other Internet subscribers, who will not be able to download a file if a complete version is not available from someone else's computer.
Persistent attempts by file-sharing software to get through may yet succeed after several minutes, as experienced in the AP's test last year. Gummadi's test did not look at the duration of the traffic blocks. Comcast has said that it is "delaying" file transfers rather than blocking them.
Robb Topolski, a former Intel Corp. engineer who noticed blocking on his home Comcast connection last year and brought it to notice, said Gummadi's work was the most authoritative study so far of this type of traffic interference. The methodology "covered all the bases," he said.
BitTorrent is one of the most popular ways to trade files online. It's commonly used to illegally share copyright material, but companies have sprung up to use it and related technologies as a cheap way to spread legal files.
Gummadi said the test did not conclusively show that Cox or Comcast were blocking traffic, since neither company carried data all the way from their subscribers to his servers in Germany. In theory, intervening carriers could be disrupting the traffic, but there is no reason to believe that they're doing so.
Apart from Comcast and Cox, Gummadi found signs of interference at seven other U.S. ISPs, all of them cable companies. But the number of blocked connections was too low to conclusively say their subscribers are being targeted, and Gummadi withheld the names of the ISPs.
StarHub is Singapore's dominant cable company. It did not reply to an e-mail to its press office.
Gummadi, who is the head of the network systems research group at the German institute, found no signs of interference by phone companies. Their DSL connections aren't shared between neighbors the same way cable is, so they have less need to manage congestion.
The percentage of blocked connections for Comcast and Cox subscribers did not appear to correspond to periods of high congestion. Subscribers were roughly equally likely to be blocked at all times of day.
Gummadi's research is aimed at figuring out the workings of Internet service providers, who are generally reluctant to share details.
"When you go to the access systems, it's almost like they're black boxes," he said. That means that programmers developing software for these systems often don't know how well it will work, he added.