- Services & Solutions
- Clients & Partners
Continuous User Authentication through Natural Speech and Language Activity
Secure user access to information systems is crucial for government and commercial enterprises. As passwords have become more and more complex, however, they are hard or impossible to remember, and are a known security risk. SRI is developing software that can uniquely identify a system’s users from spoken and written inputs, rather than using passwords.
The most secure authentication schemes are based on who a person is and how they act in everyday environments. For the Defense Advanced Research Projects Agency (DARPA) Active Authentication program, SRI research linguists are developing alternate means of secure user identification based on their natural behavioral, cognitive, and physical attributes. The result is LinguaKey—a software package that authenticates users based on their spoken and written inputs.
LinguaKey speech software goes beyond speech recognition, speech biometrics, and text stylistics to tap into core cognitive and behavioral traits in language usage. Language use reflects many factors, including personality, education, gender, native language, age, ethnicity, etc. LinguaKey uses sociolinguistics research that has defined factors that characterize different social groups as inputs to machine learning algorithms to authenticate individuals.
These same features and learning techniques can also be used to identify demographic groups in social media, identify “spear phishing” attempts by detecting fraudulent emails using a profile of a user’s real email messages, and detect network intrusions by modeling a system administrator’s typical behavior in working on information systems.