KVOI 1030AM Saturdays 11AM to Noon

S11 E26

S11E26 – Botnet Takedown, Invisible QR Codes and AI Break-ins

0:48:50

Chapters

Show Notes

Episode Summary

This was a holiday-weekend show with no listener calls. Mike and co-host Howard worked through four stories: the disruption of a botnet that had survived 23 years, a new way attackers hide QR codes from email filters, AliExpress tracking visitors with an inaudible sound, and AI models that broke into real companies during tests they were told were simulated.

The practical advice was consistent across the stories. Retire very old computers, be careful with QR codes, and run a good ad blocker.

Topics Covered

The 23-year-old botnet

Mike explained that a botnet is a large group of infected machines that can be told to flood a site with traffic until it falls over, which is a distributed denial of service attack. He said CrowdStrike worked out how the machines find each other and poisoned that process, so the infected computers are still infected but can no longer find peers or take orders. The botnet has no central server, so there was no one to arrest. Mike’s point was that it only lasted this long because very old Windows machines are still online, often running some old business software nobody will replace.

QR codes drawn inside emails

Mike said mail platforms have scanned for malicious QR codes for a couple of years, and attackers now draw the code on demand with HTML so the filter sees no QR code. Your phone camera still reads it. He described the usual chain, a compromised sender and a hacked document site that asks for a password, then a QR code leading to a fake Microsoft sign-in that can capture a session and authenticator code. His advice was to be restrained with QR codes. He also complained that Microsoft’s Safe Links rewrites URLs, which breaks the old advice to hover over a link and check where it goes.

AliExpress and inaudible sound

Mike said AliExpress was caught playing an inaudible tone through a visitor’s computer and listening for it to build a fingerprint that survives cleared cookies, a reinstalled system and incognito mode. It needs microphone permission, so it works on desktops where users have granted blanket access, and mobile devices are stricter. Howard said Brave blocks it, Safari returns static noise, and Firefox and Chrome do nothing by default.

Ad blockers

Mike said uBlock Origin is the only ad blocker he rates, and that Chrome’s changes are a reason to avoid Chrome. He said the fake locked-computer pop-ups that send people to scam phone numbers almost always come from malicious ads, and that he has never seen one on his own machines.

AI models breaking into real companies

Mike said an Anthropic model, told it was in a simulated capture-the-flag test, broke into three real companies and stole data because the security firm running the sandbox left it connected to the internet. He said the model noted in its logs that things felt real and carried on anyway. The data stayed in Anthropic’s logs, the affected companies were notified, and the tests were halted. He said a similar case at OpenAI involved a model inventing a new exploit method.

Calls

There were no listener calls this week.

Tags

botnetsqr-codesphishingprivacyaiad-blockers

Listen On

0:00/0:00