Researchers from the UC Berkeley will present a paper Wednesday at the 22nd USENIX Security Symposium in Washington, D.C., on Wednesday that explores the underground market of spam and abuse on Twitter.
Led by Vern Paxson of International Computer Science Institute (ICSI) and Chris Grier of UC Berkeley, the group tracked the criminal market on Twitter, which sells access to accounts that are later used to push spam, malicious links (including Phishing and malware), as well as inflate follower counts.
Their research took ten months, and during that time they examined 27 merchants responsible for several million fraudulent accounts. Of those, 95 percent of them were taken offline after the researchers reported them to Twitter. However, the paper says that these merchants were responsible for nearly 10-20 percent of all the illegitimate accounts created on the service during the monitoring period, and that the criminals controlling these market places earned $127,000 to $159,000 for their efforts.
The study was limited to Twitter, only because the researchers couldn't obtain permission from other social networks, such as Facebook, Google, and Yahoo -- each of which the researchers observed being actively abused by the merchants peddling accounts. As part of their service offerings, the merchants observed during the study promise such things as spam hosting, CAPTCHA solving, PPI (Pay-Per-Install) programs, and exploit kits.
"Revenue generated by miscreants participating in this market varies widely based on business strategy, with spam affiliate programs generating $12$92 million and fake anti-virus scammers $5-116 million over the course of their operations," the paper presented by the researchers notes.
"Specialization within this ecosystem is the norm," the paper goes on to explain, "The appearance of account merchants is yet another specialization where sellers enable other miscreants to penetrate walled garden services, while at the same time abstracting away the complexities of CAPTCHA solving, acquiring unique emails, and dodging IP blacklisting. These accounts can then be used for a multitude of activities..."
Mostly, the fraudulent social media profiles are used to propagate spam and Phishing attacks, but malware is a key function as well. Monetizing spam relies on grey-market and legitimate affiliate programs, the researchers discovered, as well as ad-syndication servers, and ad-based URL shortening services (example: adf.ly).
"Apart from for-profit activities, miscreants have also leveraged fraudulent accounts to launch attacks from within Twitter for the express purposes of censoring political speech. All of these examples serve to illustrate the deleterious effect that fraudulent accounts have on social networks and user safety," the researchers wrote.
According to the paper, the average cost for a Twitter account is only $0.04. Facebook accounts vary, averaging between $0.45-1.50 per account if it is phone verified, or as low as $0.10 per account without verification. Phone verified Google accounts are about $0.03-0.50 per account; while Hotmail ($0.004-0.03) and Yahoo ($0.006-0.015) accounts are priced way below the norm due to their wide availability.
"Prices ranged in price from $.10 $.15 per verification for bulk orders of 100,000 verifications, and $.25 per verification for smaller orders," the researchers add, showing a clear business plan by the merchants to move as many high-value accounts as possible in a single order.
To keep many of the accounts sold from being flagged instantly, the researchers explained that most of them are pre-aged, allowing them to avoid heuristics that disable new accounts based on weak, early signs of misbehavior. This also helps the accounts last, as older accounts must meet a much higher threshold before being nixed due to potential malicious actions.
When orders were placed for accounts, the researchers noted that many of the merchants delivered 70 percent of the promised volume within a day, and 90 percent within three days, showing a solid turnaround.
"Web services that rely on automation barriers must strike a tenuous balance between promoting user growth and preventing the proliferation of fraudulent accounts and spam behavior," the paper concludes.
"While we draw many of our observations from the Twitter account abuse problem, we believe our recommendations should generalize across web services."
One of the offered recommendations includes making fraudulent account operations more expensive to operate. One way is to focus on email verification, which raises the cost of the Twitter operation by 56 percent. After that, CAPTCHAs are another source of frustration.
"In our experience, when required, CAPTCHAs prevent merchants from registering 92% of fraudulent accounts. Services could also leverage this failure rate as a signal for blacklisting an IP address in real-time, cutting into the number of accounts merchants can register from a single IP," the researchers observed.
The paper also outlined several other recommendations, including a pattern recognition framework, which is what helped the researchers (and Twitter) retroactively identify millions of fraudulent accounts as previously mentioned.