Continuous Authentication against Collusion Attacks.

Area of Science:

• Computer Science
• Cybersecurity
• Biometrics

Background:

• Mobile devices store sensitive data, increasing the need for robust security.
• Gait authentication on smartphones offers convenience but faces security challenges.
• Smartphone sensors are increasingly sophisticated, enabling new authentication methods.

Purpose of the Study:

• To propose and validate a novel collusion attack model against smartphone gait authentication.
• To develop and evaluate a multi-cycle defense model to mitigate collusion attacks.
• To enhance the security and robustness of mobile device authentication systems.

Main Methods:

• Investigated imitation attacks and developed a collusion attack model with quantified action specifications.
• Utilized acceleration sensor data for attack simulation and defense model training.
• Implemented and tested a multi-cycle defense model based on acceleration direction changes.

Main Results:

• The proposed collusion attack successfully increased the attacker's false match rate using only an acceleration sensor in certain systems.
• The multi-cycle defense model significantly reduced the attacker's success rate.
• The defense model demonstrated improved robustness for smartphone-based gait authentication.

Conclusions:

• Gait authentication systems are vulnerable to novel collusion attacks exploiting sensor data.
• A multi-cycle defense model effectively counters collusion attacks, enhancing mobile security.
• Further research is needed to refine and implement advanced defense mechanisms for biometric authentication.