"How do I know that the new installed app behaves as described?" asks Andreas Zeller, professor of software engineering at Saarland University. So far experts have identified so-called malicious apps by checking their behavior against patterns of known attacks. "But what if the attack is brand-new?" asks Zeller.
His group seems to have found a new method to answer all these questions. Zeller summarizes the basic idea as follows: "Apps whose functionality is described in the app store should behave accordingly. If that is not the case, they are suspect."
His research group has named the software based on this idea "Chabada". For every app, it analyzes the description of its functionality that can be read in the app store. With methods from natural language processing, it identifies the main topics, for example "music". After that, Chabada clusters applications by related topics. For instance, the cluster "travel" consists of all apps that deal with traveling in some way. Using program analysis, Chabada detects which data and services are accessed by the apps. Travel apps normally access the current location and a server to load a map. So a travel app secretly sending text messages is suspicious.
The researchers applied this approach on 22,521 apps from the Google Play Store. With a purpose-built script, they had downloaded the 150 most popular apps in the 30 categories from Google Play during spring and winter of last year. Chabada then analyzed them. Finally, the computer scientists from Saarbruecken investigated the 160 most significant outliers to verify Chabada's selection. The result: It had detected 56 percent of the existing spy apps, without knowing their behavior patterns beforehand.
How important the researchers' efforts are is shown by a news item published by the Russian software company "Doctor Web" at the end of June last year. It reported that the company had discovered various malicious apps on the "Google Play" platform. Downloaded onto a smartphone, the malware installed other programs, which secretly sent text messages to expensive premium services. Although Doctor Web, according to its own statement, informed Google immediately, the malicious apps were still available for download for several days. Doctor Web estimates that in this way up to 25,000 smartphones were used fraudulently. "In the future Chabada could serve as a kind of gatekeeper, ensuring that malicious apps will never make it into an app store", Zeller explains.
The computer scientists from Saarbruecken will present their new approach at the International Conference on Software Engineering (ICSE) in Hyderabad, India at the end of May. Already in March, Google security researchers will be meeting with the Saarbruecken team. Google has also already invited Zeller and his colleagues to have Chabada analyze the whole Google App Store.
Background of computer science in Saarbruecken at Saarland University:
The core of computer science in Saarbruecken is the faculty of computer science. Only a few meters away on campus are seven more internationally well-respected research institutes. Apart from the two Max Planck Institutes, for Informatics and Software Systems, these are the German Research Center for Artificial Intelligence (DFKI), the Center for Bioinformatics, the Center for IT-Security, Privacy and Accountability (CISPA) and the Cluster of Excellence "Multimodal Computing and Interaction".
Gordon Bolduan Science Communication Competence Center Computer Science Saarland Email: firstname.lastname@example.org Tel: +49 681 302 70741
Tips for radio journalists:
You can do telephone interviews in studio quality with Saarland University researchers via radio codec (IP connection with direct dialing or via ARD-star-dot 106813020001). Please contact the editorial department for interview requests (+49 681/302-3610).