Researchers have developed an adversarial machine learning algorithm to improve detection of malicious JavaScript code by using large language models to rewrite it. Unlike off-the-shelf obfuscation tools, which create obvious changes that can be detected, the tool creates changes that look natural and are harder to detect. Retraining deep learning-based detectors on adversarially generated samples improved their performance by 10%. The researchers warn that while detecting malware becomes challenging as it evolves, using similar tactics to rewrite such code can aid in improving machine learning models.
When insider risk is a wellbeing issue, not just a disciplinary one
Written by Katie Barnett, Director of Cyber Security at Toro Solutions Insider risk is still often framed around intent, with the focus placed on malicious
