Endpoint
3/17/2010
05:23 PM
Connect Directly
Google+
Twitter
RSS
E-Mail
50%
50%
Repost This

New Free Service Filters Twitter Spam

Spamdetector sniffs out spammer accounts, which Twitter then deletes

Researchers recently launched a free spam-filtering service for Twitter that flags offenders for the microblogging service, which, in turn, removes the offending accounts.

The Spamdetector service crawls Twitter, using heuristics to detect spam accounts. Gianluca Stringhini, the researcher heading up the service, says the tool has a low false-positive rate.

"Around a year ago, we started observing these [social] networks, looking for malicious traffic...spammers show a very different behavior compared to real users, and therefore we were able to build a system that can detect them in a reliable way," says Stringhini, a researcher at the University of California-Santa Barbara.

Twitter already offers its users a button for reporting possible spam, but it relies on the user to spot suspicious tweets. The social network last week also added a new service that detects malicious URLs in an effort to quell the rise in spam and phishing on the network. It ultimately will scan all URLs before they hit the Twitter feed, but initially is doing so only for URLs sent via Twitter direct messages [DMs] and email notifications about DMs.

To be sure, abuse on Twitter is on the rise: One in eight Twitter accounts was found to be suspicious, malicious, or suspended last year, according to a recent report from Barracuda Networks. And Twitter itself currently finds 3 to 4 percent of accounts to be malicious, according to Barracuda's data.

Spamdetector also takes potential spam submissions from users who want to check if a tweet is legitimate. "Our system crawls Twitter on its own, and the spam detection happens mostly without users' submissions. We ask users to flag spammers to us because, in this way, we can target the crawling to certain accounts -- for instance, those sending tweets that have been detected as spammers -- allowing us to detect more spammers in a shorter period of time," Stringhini says. "When we detect spammers, we flag them to Twitter, which then takes care of deleting them."

Spamdetector profiles spammers based on features that are common in spammer accounts. "For example, spammers usually follow a large number of profiles, but are followed back only by a small fraction of them. They also usually send out very similar messages," Stringhini says. "We calculate some ratios related to the defined features for each account and flag as spammers those profiles that exceed a certain amount of thresholds in those ratios."

David Maynor, CTO of Errata Security, which runs an experimental service for detecting Twitter spam and threats called TwiGUARD, says spam has grown 3 percent since TwiGUARD first launched in September.

TwiGUARD was one of the first security tools for Twitter. It uses a reputation-based algorithm to see if a follower is a spammer or if a link is infected. The service also analyzes malware found via tweets. Maynor says Errata could eventually offer TwiGUARD as a full-blown service. "We're fine-tuning the algorithms for developing a user-friendly front-end to it," he says.

UC-Santa Barbara's Stringhini, meanwhile, says Twitter is actively fighting spam and malicious accounts on its network. "It is not easy to detect malicious accounts among millions of users. Moreover, Twitter is peculiar because many legitimate accounts are managed in an automated fashion, [such as] profiles of shops and posting deals, and this might lead to a wrong detection," he says. But Twitter is having some success on its own -- some have already been suspended when Spamdetector has flagged them, for example, he says.

As of Tuesday, the new service had detected -- and removed -- nearly 3,400 spam bot accounts in about a month.

To get the Spamdetector services, just follow @spamdetector on Twitter.

Have a comment on this story? Please click "Discuss" below. If you'd like to contact Dark Reading's editors directly, send us a message. Kelly Jackson Higgins is Senior Editor at DarkReading.com. She is an award-winning veteran technology and business journalist with more than two decades of experience in reporting and editing for various publications, including Network Computing, Secure Enterprise Magazine, ... View Full Bio

Comment  | 
Print  | 
More Insights
Register for Dark Reading Newsletters
White Papers
Cartoon
Current Issue
Video
Slideshows
Twitter Feed
Dark Reading - Bug Report
Bug Report
Enterprise Vulnerabilities
From DHS/US-CERT's National Vulnerability Database
CVE-2012-3946
Published: 2014-04-24
Cisco IOS before 15.3(2)S allows remote attackers to bypass interface ACL restrictions in opportunistic circumstances by sending IPv6 packets in an unspecified scenario in which expected packet drops do not occur for "a small percentage" of the packets, aka Bug ID CSCty73682.

CVE-2012-5723
Published: 2014-04-24
Cisco ASR 1000 devices with software before 3.8S, when BDI routing is enabled, allow remote attackers to cause a denial of service (device reload) via crafted (1) broadcast or (2) multicast ICMP packets with fragmentation, aka Bug ID CSCub55948.

CVE-2013-6738
Published: 2014-04-24
Cross-site scripting (XSS) vulnerability in IBM SmartCloud Analytics Log Analysis 1.1 and 1.2 before 1.2.0.0-CSI-SCALA-IF0003 allows remote attackers to inject arbitrary web script or HTML via an invalid query parameter in a response from an OAuth authorization endpoint.

CVE-2014-0188
Published: 2014-04-24
The openshift-origin-broker in Red Hat OpenShift Enterprise 2.0.5, 1.2.7, and earlier does not properly handle authentication requests from the remote-user auth plugin, which allows remote attackers to bypass authentication and impersonate arbitrary users via the X-Remote-User header in a request to...

CVE-2014-2391
Published: 2014-04-24
The password recovery service in Open-Xchange AppSuite before 7.2.2-rev20, 7.4.1 before 7.4.1-rev11, and 7.4.2 before 7.4.2-rev13 makes an improper decision about the sensitivity of a string representing a previously used but currently invalid password, which allows remote attackers to obtain potent...

Best of the Web