CASE STUDY

Developing new AI techniques for securing mobile applications

US Department of Defense partners with MIT and Blu Sphinx to develop a new approach to secure mobile applications

 

There are hundreds of thousands of US Department of Defense (DoD) employees and service personnel who require secure access to data on mobile devices. However, today it is very hard to know if a malicious application is on a mobile device. DoD was looking for a novel technology to help secure data and find malicious applications on mobile Android devices.

Client
US Department of Defense
Year

2017

Challenges
  • Inadequate cyber security
  • Finding malware
  • Many mobile applications
Outcomes
  • Can use mobile applications
  • New way to detect malware
  • Increased cyber security
The Main Problem

Human analysis and traditional security tools are inadequate to solve this problem

DoD spent years trying to develop a process to vet mobile applications and monitor their trustworthiness. Human analysis is very slow and unreliable at detecting malicious applications and traditional security approaches can be circumvented. DoD had made significant investments into standard security technologies, such as malware signature detection and various dynamic monitoring technologies, but none was able to provide the desired security level for their mobile devices.

 Despite being a very expensive and time-consuming exercise, this effort had proven insufficient at stopping malware applications from being installed at DoD mobile devices. Consequently, potentially extremely sensitive details, such as the GPS coordinates of a soldier in the field, could be accessed by adversaries.

The Outcome

Blu Sphinx AI technology helped model all possible behaviors of mobile applications to detect hidden malicious code

In partnership with MIT, our Blu Sphinx AI technology was used to create a formal model of a mobile application from its code, and then to check if the created application model exhibits any potential malware behaviors.

 By creating a formal model of an application and its mobile operating systems functions, the problem of identifying malware in a mobile android application can be mapped to finding a possible solution to a system of constraints, such as “is it possible that my location information flows to anyone outside my trusted network?” In doing so, checking for any hidden malware becomes a constraint solving problem.

 An evaluation of the project at the end demonstrated that this technology achieved unprecedented precision and accuracy for the information-flow analysis of Android mobile applications. This technology can detect malicious sensitive information leaks inserted by sophisticated independent hostile organizations, where a current state-of-the-art analysis largely fails.

What the project leader says about the Blu Sphinx impact

“Dr. Dillig’s research produced multiple internationally known results, published in the top publication outlets, and established his position at the very top of the field.

One of Dr. Dillig’s notable results centers around the formulation of complex software security properties using satisfiability modulo theories (SMT) solvers. He applied this approach to the problem of finding certain classes of security vulnerabilities.

It moves from simply pointing out the presence of errors to proactively taking actions to eliminate errors. My research group at MIT had obtained a sizable contract. I found [him] to be a model collaborator. His ability to work productively with the team and to communicate sophisticated technical concepts quickly and efficiently contributed greatly to our ability to execute on the contract.

In large part because of Dr. Dillig’s participation, the project was a success and we were very happy with the collaboration.”

Professor Martin Rinard ,PhDProfessor, MIT Computer Science and Artificial Intelligence Laboratory

Ready to start working on a smarter ERP? Contact us for a free analysis and demo.