acm-header
Sign In

Communications of the ACM

ACM Opinion

A Developer's Guide to Machine-Learning Security


Illustration showing adversarial attacks on machine learning app, changing descriptions of images

The first step to countering adversarial attacks is to understand the different types and the weak spots of the machine-learning pipeline.

The threat of adversarial attacks has become one of the important concerns of machine-learning (ML) applications. Adversarial attacks are different from other types of security threats that programmers are used to dealing with.

Machine learning needs new perspectives on security. Developers must learn to adjust their software development practices according to the emerging threats of deep learning as it becomes an increasingly important part of their applications.

From TechTalks
View Full Article


 

No entries found