17 August 2017. Computer scientists developed techniques that help identify the people placing sexual service advertisements by tracing the bitcoin payments for those ads. The technology is described in a paper delivered this week at the Conference on Knowledge Discovery and Data Mining in Halifax, Nova Scotia, by a team from University of California in Berkeley and San Diego, as well as New York University.
The scale of forced sexual exploitation is global and substantial. International Labor Organization estimates 4.5 million individuals worldwide are forced into commercial sex services, about 1 in 5 of all 21 million people forced into slavery-like working conditions. In the U.S. the National Center for Missing and Exploited Children says in 2016 1 in 6 runaway children, all under the age of 18, were forced into prostitution rings, including girls, boys, and LGBTQ youth.
Authorities seeking to break these sexual exploitation businesses are often stymied by the anonymous nature of their transactions, even when the traffickers advertise openly in online publications. Among the ways traffickers evade detection is by paying for the ads with bitcoins, a private peer-to-peer exchange technology that enables funds transfers between anonymous accounts, which online publications accepted for sex ads at the time of the study. Bitcoin is based on blockchain, a technology linking online transactional ledgers or registries exchanging encrypted blocks of data that establish a form of trust between the parties.
The team led by UC-Berkeley doctoral candidate Rebecca Portnoff developed algorithms to unmask individuals placing the ads in online publications for so-called adult services. One algorithm analyzes text in the ads using a technique called stylometry that statistically measures factors such as sentence length and word frequency to reveal writing style signatures to identify, for example, plagiarists. In this case, the researchers employed stylometry in a machine learning routine to highlight ads written by the same person.
A second set of algorithms finds and exploits leaks in the bitcoin exchange system to trace financial transactions paying for the ads. In this case, the leaks are time stamps on transactions appearing in an intermediate public bitcoin confirmation mechanism called mempool. The researchers discovered that Backpage, one of the leading online media for escort and other prostitution services, posted new ads at almost the exact time of the transaction reaching the mempool. By matching the time stamps of Backpage ads with the mempool, the team could possibly match the bitcoin transaction identifiers to the Backpage ads.
Portnoff and colleagues then devised techniques for matching ad buyers with the time stamps of the ads and mempool appearance. One method the team calls “hard identifiers” maps the bitcoin wallet, or account, identifier to individual ad with the fewest number of nodes crossed, no more than 2, to get from the wallet to the ad. Another technique the researchers call “persistent bitcoin identity” tracks bitcoin traffic linked to the ads, when the wallet identifier is adjacent to the ad purchasing transaction.
To test their techniques, the team placed a series of phony escort ads in Backpage that enabled the researchers to trace transactions they initiated to ads appearing in the publication by matching the respective time stamps. The researchers also collected a month’s worth of adult ads from Backpage, some 10,000 in all, where they found the stylometric algorithm accurately found the same author nearly 90 percent of the time. However, the team could not consistently identify individual ad buyers to individual bitcoin transactions, but the ability to highlight suspects and narrow the search could benefit law enforcement authorities and non-government organizations fighting human trafficking.
Since the study, Backpage removed its adult advertising section, but the researchers say that ads for sex services still appear in other sections of the site.
More from Science & Enterprise:
- Creative Ideas Mined with Crowdsourcing, Neural Nets
- Life Science Execs: Financial, IP Info Top Hacker Targets
- Start-Up Applying Artificial Intelligence to New Drug Uses
- Software to Simplify Big Data Curation
- Ford Investing $1B in Artificial Intelligence Start-Up
* * *
You must be logged in to post a comment.