Facebook Discusses Fight Against Spam
Facebook offered users a glimpse at its anti-spam arsenal recently in an effort to be more transparent.
With billions of pieces of content being shared on Facebook every month and spammers relentlessly targeting users, fighting spam isn't easy, blogged Caroline Ghiossi, an associate on Facebook's user operations team. The site relies in part on its users to report bad behavior, and has used that information to identify common patterns of unacceptable behavior, she added.
"For example, we've learned that if someone sends the same message to 50 people not on his or her friend list in the span of an hour, it's usually spam," she wrote. "Similarly, if 75 percent of the friend requests a person sends are ignored, it's very likely that that person is annoying others he or she doesn't actually know."
Without going into detail about how the systems work, she noted that the anti-spam systems are designed to detect suspicious behavior and block it.
"In extreme cases where the behavior continues despite our warnings, we may disable the person's account," she wrote. "When this happens, it usually isn't a person's account at all but a fake account or a real account that's been compromised. The compromised accounts are put into a process to give control back to the rightful owner. In all other cases, we always give the person an opportunity to appeal the decision by contacting us."
Facebook's automated systems also protect against malicious Websites by adding them to block lists and deleting Wall posts or messages that link to it.
"Sometimes, spammers try to hide their malicious links behind URL shorteners like Tiny URL or bit.ly, and in rare cases, we may temporarily block all use of a specific shortener," she added. "If you hit a block while using a URL shortener, try a different one or just use the original URL for whatever you're trying to share."
"These systems are so effective at working in the background that most people who use Facebook will never encounter one," she continued. "They're not perfect, though, and we're always working to improve them. We do this by actively monitoring appeals and learning from the rare cases in which we make mistakes."