cognitive cybersecurity intelligence

News and Analysis

Search

Framework to Detect Backdoor Attacks on Deep Models

DeBackdoor is a new framework designed to detect stealthy backdoor attacks on deep learning models before deployment. It functions under real-world constraints, requiring limited data and only black-box access. Utilizing a Simulated Annealing optimization algorithm, DeBackdoor effectively identifies various attack types, outperforming existing methods and enhancing security for safety-critical applications.

Source: cybersecuritynews.com –

Subscribe to newsletter

Subscribe to HEAL Security Dispatch for the latest healthcare cybersecurity news and analysis.

More Posts

What are business logic vulnerabilities?

Business logic vulnerabilities in software allow attackers to exploit flaws in design, enabling them to circumvent security measures and manipulate pricing, authentication, and other key