A Collection of Articles
Edit

Computing

Tracking Social Spam

Hacked Facebook accounts produce most of the spam on the world’s largest social network.

Source: “Detecting and Characterizing Social Spam Campaigns”
Hongyu Gao et al.
ACM Internet Measurement Conference, November 1-3, 2010, Melbourne, Australia

Results: In the first large-scale study of spam activity on Facebook, researchers at Northwestern University and the University of California, Santa Barbara, showed that most of these messages originate from compromised accounts rather than from phony profiles. Spammers use those accounts to send messages to a victim’s friends. Typically, the messages contain links to pharmacy sites or pages that attempt to steal more account details.

Why it matters: With more than half a billion users, Facebook is a tempting target for spammers but one that is much more difficult to exploit than e-mail. Understanding how spammers are using Facebook is important to developing defenses against them as the network continues to grow and as attack strategies evolve. Academics say that publicly reporting spam also encourages Facebook to be more forthcoming about its efforts to protect users from this activity.

Methods: Facebook is built around the concept of letting you share information only with other users you have chosen to connect with. But until late 2009, most users were part of regional “networks” that allowed all the people in, say, a given city or university to access one another’s profiles. The researchers made use of that feature and joined several large regional networks. They then downloaded the profiles and public messages of roughly 3.5 million users. The public messages obtained were analyzed for traces of spam activity.

Next steps: The researchers plan to share their results with Facebook to help the company reduce spam and identify telltale patterns that could be used to spot accounts taken over by spammers. They will also continue to run similar analyses to track how spam activity and strategies change.

Fast Processing

A new algorithm could speed image processing, recommendation systems, and more

Source: “Approaching Optimality for Solving SDD Linear Systems”
Ioannis Koutis et al.
IEEE Symposium on Foundations of Computer Science, October 23-26, 2010, Las Vegas, Nevada

Results: A new algorithm offers a significantly faster way to solve certain systems of linear equations. The Carnegie Mellon researchers who created the algorithm say it could enable an ordinary desktop computer to solve billion-­variable systems in seconds.

Why it matters: The equations that this algorithm can solve have a wide variety of practical applications. For example, Netflix uses them to factor in myriad variables in making movie recommendations. In image processing, they are used to identify different parts of a picture and remove blurry spots. They are also used to optimize systems—calculating, for example, the maximum number of vehicles that can run through a network of highways.

Methods: To solve a complex set of equations, researchers typically start by producing a simplified version that is easily solved. This version can guide them in tuning the steps to solving the full system of equations. In the new study, the researchers used a combination of graph theory techniques to come up with a much better way to simplify the system.

Next steps: The algorithm can be used to create new “solvers”—series of algorithms that can solve for the variables in a given system of linear equations. The researchers are working to extend their methods to different types of linear-equation systems, which would allow them to solve more types of real-world problems more efficiently.

Uh oh–you've read all five of your free articles for this month.

Insider basic

$29.95/yr US PRICE

Subscribe
What's Included
  • 1 year (6 issues) of MIT Technology Review magazine in print OR digital format
  • Access to the entire online story archive: 1997-present
  • Special discounts to select partners
  • Discounts to our events

You've read of free articles this month.