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:
- A Comparative Study on Linux and Windows Ransomware Attacks
November 21, 2023
During the last few months, CheckPoint researchers conducted a study of some of the top ransomware families (12 in total) that either directly developed ransomware for Linux systems or were developed in languages with a strong cross-platform component, such as Golang or Rust, thereby allowing them to be compiled for both Windows and Linux indiscriminately. The ...
- When Maximum Effort Doesn’t Equate to Maximum Results
November 21, 2023
It’s no secret that security teams are feeling beleaguered as a result of the barrage of data, events, and alerts generated by their security tools, to say nothing of the increased budget scrutiny and constrained staff resources that continue to plague cybersecurity practitioners. The trick is finding the right balance between how much internal teams have ...
- Konni Campaign Distributed Via Malicious Document
November 20, 2023
FortiGuard Labs recently identified the use of a Russian-language Word document equipped with a malicious macro in the ongoing Konni campaign. Despite the document’s creation date of September, ongoing activity on the campaign’s C2 server is evident in internal telemetry. This campaign relies on a remote access trojan (RAT) capable of extracting information and executing commands ...
- Social engineering attacks lure Indian users to install Android banking trojans
November 20, 2023
Microsoft has observed ongoing activity from mobile banking trojan campaigns targeting users in India with social media messages designed to steal users’ information for financial fraud. Using social media platforms like WhatsApp and Telegram, attackers are sending messages designed to lure users into installing a malicious app on their mobile device by impersonating legitimate organizations, such ...
- Hackers accessed sensitive health data of Welltok patients
November 20, 2023
Hackers accessed the personal data of more than a million people by exploiting a security vulnerability in a file transfer tool used by Welltok, the healthcare platform owned by Virgin Pulse. Welltok, a Denver-based patient engagement company that works with healthcare plans to provide communications to subscribers about their healthcare, confirmed in a data breach notification ...
- K-12 schools improve protection against online attacks, but many are vulnerable to ransomware gangs
November 19, 2023
Some K-12 public schools are racing to improve protection against the threat of online attacks, but lax cybersecurity means thousands of others are vulnerable to ransomware gangs that can steal confidential data and disrupt operations. Since a White House conference in August on ransomware threats, dozens of school districts have signed up for free cybersecurity services, ...