ポッドキャスト

#9 - The collapse of LAPSUS$ and the risks of AI data poisoning

Omer Mosheiov

9月 13, 2023

聞いてくださいSpotify

聞いてくださいApple Podcasts

👀 Here's a sneak peek at today’s episode:

🔒 Stay ahead of the game! LAPSUS$ Hackers may be making waves. Two members of this notorious group faced consequences in the UK, but shockingly, they continued their hacking activities even while under house arrest.

🤖 Data Poisoning in AI Training is a growing concern. Hackers can manipulate the data used to train AI models, introducing risks and vulnerabilities. Validating data integrity and randomizing data ingestion times are useful mitigations against this threat.

💻 The WinRAR Vulnerability (CVE-2023-38831)! This flaw was exploited against crypto-traders to infect their devices with malware, but should be considered a low concern for cloud customers unless using virtual desktops.

リソース

https://gizmodo.com/hackers-lapsus-uber-nvidia-rockstar-games-microsoft-1850766324 https://www.bbc.com/news/technology-66549159 https://www.cisa.gov/resources-tools/resources/review-attacks-associated-lapsus-and-related-threat-groups-executive-summary https://www.cisa.gov/sites/default/files/2023-08/CSRBLapsus%24508c.pdf https://duo.com/decipher/lapsususd-analysis-finds-need-for-better-iam-mfa-deployments https://www.youtube.com/watch?v=h9jf1ikcGyk https://arxiv.org/pdf/2302.10149.pdf https://www.blackhat.com/us-23/briefings/schedule/#poisoning-web-scale-training-datasets-is-practical-32112 https://arstechnica.com/security/2023/08/winrar-0-day-that-uses-poisoned-jpg-and-txt-files-under-exploit-since-april/