TechLetters ☕️Loud coverage of disinformation/propaganda helps propagandists. Cognitive operation using faked ChatGPT error messages. Hacking nuclear reactors, vacuum cleaners, Russian media; and more
Overhyping disinformation considered harmful
How to boost Russian propaganda? The SDA, Disinformation Factory, employed over 100 staff. Between November 2023 and August 2024, its Doppelganger project produced 700+ fake websites, becoming one of Russia's largest disinformation factories. Leaked documents show forgery is central to Russia's strategy, echoing Soviet-era methods. Doppelganger's exposure in media, with 163 stories, inadvertently boosted its perceived impact. "Exposing digital disinformation products has become its own cottage industry. Dozens of nonprofit and for-profit outfits now focus on hunting for the next influence network to expose with as much fanfare as possible, no matter how insignificant the disinformation projects might be. But such surface-level downstream exposure no longer deters adversaries. In fact, it helps them get more funding". In other words: most disinformation/propaganda is insignificant (non-event), overhyping it might profit the propagandists, and some who voice it.
Now, what do we do about it? We wait for the next article/assessment debunking no0-impact narratives put to unpopular and unvisited blogs, sites, social media accounts with 3 followers, of course fact-checking any such worthless content?
Security
Remember this? Confirmed: It was a hoax/troll!
It did not originate from GPT system, and contained errors. "non-AI activity was used to deceive people about the use of our models". But it did gather more viewers to other questionable content in this account. That is an example of a cognitive operation: lots of people assumed that ‘bots’ are aplenty and that they use AI. So here someone played on this broad expectation. More about such operational tactics and techniques in my book Propaganda.
U.S. indicts three suspected for attempts to interfere in the 2024 U.S. election. They broke in Trump’s campaign staff accounts, stolen information, and leaked it to the media and Biden’s campaign staff. They defeated two-factor authentication.
Poorly designed connected car systems are a big risk. As demonstrated recently with the disclosure of fatal design of KIA car management systems. Only knowing a number plate, hackers or Bad People could control any car all over the world. And this is the reality we are creating for ourselves.
Major Russian state media were down following to a cyberattack. "Online broadcasting and internal services are not working, there is no Internet or telephony". Reports of data destruction (including backups) and expectation of long down-time. Rossia 1 and Rossia 24 (TV), in addition to more than 80 regional television and radio stations
Ecovacs automatic (Chinese) vacuum cleaners hacked. Cameras built into the devices (real-time) accessed. As the vacuum cleaners have speakers, they were used to utter slurs. There goes your smart home appliance.
Hacking nuclear reactors? Hacking nuclear reactors? Technical analysis of hypothetical cyber-physical attacks targeting the safety systems of nuclear reactors (PWRs), such as the Reactor Protection System (RPS) and the Engineered Safety Features Actuation System (ESFAS). The sky is not falling, but this writeup should be reviewed more widely by security community. NPP safety works notoriously in a closed-information loop, which is supposed to be an additional security measure (?), but let's face it - those who would have the means and intent to target them would have rather effortless access to all the details imaginable. We did consider such cyber risks in the ICRC analysis and report on on the humanitarian consequences of cyber operations (description).
Privacy
Mozilla is entering the business of privacy-improved advertisement technology. Having researched and tracked web privacy & web browsers for years, I must says that this is a very interesting development!
Multi-party computation to exchange classified information about vessel discovery (radar fingerprints) between allied states. Here.
The German Competition authority closes Meta's case concerning whether combining data from different sources is legal. The Court of the EU ruled how it might be (Meta did change it to ask for consent [or pay]), so the competition authority is satisfied.
Technology Policy
Unfollowing hyperpartisan influencers has lasting positive effects on reducing out-party animosity and increasing satisfaction with Twitter feeds. The intervention also improved well-being, particularly when participants followed science accounts, while reducing exposure to political content and misinformation. Participants who re-followed hyperpartisan accounts experienced diminished long-term benefits. Overall, this low-effort strategy effectively reduces partisan animosity and offers a more targeted alternative to broader social media reduction approaches.
Other
Large language models (~AI) are not yet capable of true reasoning. Research from Apple scientists challenges their mathematical reasoning limits by introducing a benchmark designed to test LLMs on diverse mathematical problems. Even slight changes in numerical values or problem complexity can cause significant performance drops. Not capable of logical reasoning..
In case you feel it's worth it to forward this content further:
Subscribed
If you’d like to share: