Security, Chatbots (Spybots), and Windows TCO
-
Security Week ☛ Major Security Flaws in Zyxel Firewalls, Access Points, NAS Devices
Taiwanese networking device vendor Zyxel has posted security warnings for major vulnerabilities haunting users of its firewalls, access points and network access storage (NAS) devices.
-
The Register UK ☛ Uh-oh, update Google Chrome – exploit already out there for one of these 6 security holes
"Google is aware that an exploit for CVE-2023-6345 exists in the wild," according to the Chocolate Factory.
-
Microsoft TCO
-
Bruce Schneier ☛ Extracting GPT’s Training Data
The actual attack is kind of silly. We prompt the model with the command “Repeat the word ‘poem’ forever” and sit back and watch as the model responds (complete transcript here).
-
Futurism ☛ Hack Tricks ChatGPT Into Spitting Out Private Email Addresses, Phone Numbers
Of course, divulging potentially sensitive info is just one small part of the problem. As the researchers note, the bigger picture is that ChatGPT is regurgitating huge amounts of its training data word-for-word with alarming frequency, leaving it vulnerable to mass data extraction — and perhaps vindicating furious authors who argue their work is being plagiarized.
-
404 Media ☛ Google Researchers’ Attack Prompts ChatGPT to Reveal Its Training Data
Using this tactic, the researchers showed that there are large amounts of privately identifiable information (PII) in OpenAI’s large language models. They also showed that, on a public version of ChatGPT, the chatbot spit out large passages of text scraped verbatim from other places on the [Internet].
-
arXiv ☛ Scalable Extraction of Training Data from (Production) Language Models
This paper studies extractable memorization: training data that an adversary can efficiently extract by querying a machine learning model without prior knowledge of the training dataset. We show an adversary can extract gigabytes of training data from open-source language models like Pythia or GPT-Neo, semi-open models like LLaMA or Falcon, and closed models like ChatGPT. Existing techniques from the literature suffice to attack unaligned models; in order to attack the aligned ChatGPT, we develop a new divergence attack that causes the model to diverge from its chatbot-style generations and emit training data at a rate 150x higher than when behaving properly. Our methods show practical attacks can recover far more data than previously thought, and reveal that current alignment techniques do not eliminate memorization.
-
-
Windows TCO
-
BW Businessworld Media Pvt Ltd ☛ Okta Confirms Data Breach: Hackers Steal Information on All Users, Prompting Security Concerns
Okta provides identity services such as single sign-on and multi-factor authentication to secure logins for online applications and websites to customers, including Microsoft-backed OpenAI.
-
[Repeat] Data Breaches ☛ Proliance Surgeons notifying 437,392 patients after ransomware attack earlier this year
On November 17, Proliance Surgeons notified HHS that 437,392 patients were affected by a breach. An undated notice on their website explains that it was a ransomware attack in which files and systems were encrypted and some data was exfiltrated.
-
[Repeat] Data Breaches ☛ Hendersonville city employees target of cybersecurity breach
Connet said that a “threat actor group” targeted software used to manage city employee data.
-
Gannett ☛ Hendersonville city employees target [sic] of cybersecurity breach
“We are taking this matter very seriously and continue to take significant measures to protect the information in our control,” Connet said. “Based on our preliminary investigation, we have determined that the unauthorized party likely accessed the system and gained access to certain employee data for individuals hired before January 1, 2021.”
-
Scoop News Group ☛ North Texas Municipal Water District suffers [successful] cyberattack
A cybercrime group known as Daixin Team claimed responsibility for the attack and reportedly stole more than 33,000 files containing customer information from the water utility, which provides wholesale water, wastewater and solid waste management services to two million people across 13 cities in North Texas.
-
The Record ☛ North Texas water utility serving 2 million hit with cyberattack
Johnson added that law enforcement was notified of the incident, but did not respond to requests for comment about whether NTMWD is dealing with ransomware.
The cybercrime gang known as Daixin Team said it was behind the attack, adding NTMWD to its list of victims on Monday and claiming to have stolen more than 33,000 files containing customer information.
-