Plainly speaking, Artificial intelligence is no longer a fringe technology. It has become a core component of modern business, from customer service chatbots to complex data analysis.
We often treat the Large Language Models (LLMs) that are at the core of this technology as trusted black boxes. But like any software, they can be tampered with, manipulated, and turned against their creators. Understanding the ways an AI model can be compromised is the first step toward building a secure and resilient AI infrastructure. There are three primary ways that the authors want to cover, in which they will explain how an AI model can be compromised and made to act in unpredictable (or malicious!) ways:
Read more…
Source: Trend Micro
Sign up for the Cyber Security Review Newsletter
The latest cyber security news and insights delivered right to your inbox
Related:
- Financial Fraud APK Campaign
January 12, 2024
During Unit 42 ivestigation discovering threats in legitimate network traffic, activity generated by a certain type of Android Package Kit (APK) files kept hitting their radar. The research revealed a family of malicious APKs targeting Chinese users that steals victim information and conducts financial fraud. To do this, the threat actor masquerades as a law enforcement ...
- Dallas says cyberattack targeted more people than previously disclosed
January 11, 2024
Hackers who targeted the city of Dallas had access to the addresses, Social Security numbers and other personal information of nearly 300 more people than what had been previously disclosed to the public, city officials now say. The city’s spokesperson confirmed on Wednesday that further internal investigations into the cyberattack determined an additional 293 people, including ...
- Medusa Ransomware Turning Your Files into Stone
January 11, 2024
Unit 42 Threat Intelligence analysts have noticed an escalation in Medusa ransomware activities and a shift in tactics toward extortion, characterized by the introduction in early 2023 of their dedicated leak site called the Medusa Blog. Medusa threat actors use this site to disclose sensitive data from victims unwilling to comply with their ransom demands. As ...
- Weaponizing Apache OFBiz CVE-2023-51467
January 11, 2024
On December 26, SonicWall disclosed an authentication bypass affecting Apache OFBiz. SonicWall demonstrated the vulnerability, assigned CVE-2023-51467, by accessing the protected HTTP endpoint /webtools/control/ping without authentication. While that proved the vulnerability existed, it did not demonstrate arbitrary code execution. However, CVE-2023-51467 can be used to execute arbitrary code. And even better, it can be used to ...
- Distributed Energy Generation Gateway (In)Security
January 11, 2024
Distributed energy generation (DEG) is a term used to describe the shift from centralized energy generation, such as power companies, to a source — typically a renewable energy source — closer to the user. It is most often seen in the form of solar panels on the rooftops of houses or office buildings, but DEG also ...
- Actively exploited 0-days in Ivanti VPN are letting hackers backdoor networks
January 10, 2024
Unknown threat actors are actively targeting two critical zero-day vulnerabilities that allow them to bypass two-factor authentication and execute malicious code inside networks that use a widely used virtual private network appliance sold by Ivanti, researchers said Wednesday. Ivanti reported bare-bones details concerning the zero-days in posts published on Wednesday that urged customers to follow mitigation ...