A Review of Deep Learning Security and Privacy Defensive Techniques

Table 1

Comparison of attacking techniques against Deep Learning.

Attacking technique	Advantages	Disadvantages	Countermeasure technique
Causative attack [45, 46]	Influence on training data and exploits misclassifications	Time consuming	[45, 47–49]
		Not fit for large dataset
Exploratory attack [47]	Changes the discriminant results	Resource consuming	[50–52]
	Misclassifies positive sample
Integrity attack [53]	False negative passes through the system	Easily detected	[54–56]
Availability attack [57]	False positive results in blocking records	Time and resource consuming	[58–60]
Privacy violation attack [61]	Easily exploit the training dataset	Its performance is not reliable as it based on iterations	[62–64]
Targeted attack [65]	Misclassified to any arbitrary class	It does not provide assurance about the generated samples	[66–68]
Indiscriminate attack [69]	Good trade-off	Perturbation is high	[70, 71]
	Highly efficient