Table 15 Model robustness against adversarial attacks.

From: Design of a dynamic trust management and defense decision system for shared vehicle data based on blockchain and deep reinforcement learning

Attack type

Original detection rate

Enhanced model detection rate

Baseline model detection rate

Data poisoning (5%)

93.8%

92.6%

79.3%

Adversarial sample perturbation

93.8%

91.2%

68.4%

Model inversion attack

93.8%

90.8%

72.1%

Back to article page