245 - 50 Documents Can Poison AI Models - CISA KEV Adds 12 Decade-Old Vulnerabilities and Salesforce Ransomware

YusufOnSecurity.com

Enjoying the content? Let us know your feedback!

This week, we've got three stories that really caught my attention, and honestly, they're all pretty alarming in their own ways.

If you're new here, welcome to the show where we break down the latest cybersecurity news and help you understand what's really happening in the cyber security domains.

We're going to talk about a shocking discovery about AI security - turns out it takes way fewer malicious documents than anyone thought to completely poison an AI model. Then we'll discuss something that should make every security professional cringe - CISA just added a dozen vulnerabilities to their Known Exploited Vulnerabilities catalog, and half of them are over a decade old. And finally, we'll cover Salesforce's bold decision not to pay ransom to hackers who claim to have stolen data from dozens of major companies.

- https://www.anthropic.com: Small Samples Poison

- https://www.turing.ac.uk: LLMS May Be More Vulnerable Data Poisoning W Thought

- https://www.theregister.com: Salesforce Refuses To Pay Ransomware

- https://www.sans.org: CISA Adds 12 CVEs to KEV; Half are a Decade or More Old

Be sure to subscribe!
You can also stream from https://yusufonsecurity.com
In there, you will find a list of all previous episodes in there too.