AdfichterRE: https://infosec.exchange/@adfichter/116329160026076266
KEIN APRILSCHERZ
@kgndiue
- 6 Followers
- 63 Following
- 321 Posts
bert hubert πΊπ¦πͺπΊπΊπ¦After receiving the first LLM-generated pull requests, I have decided to blanket no longer look at those. When studying a PR, I take into account who made it, and if they've previously been careful developers. LLM-generated code I have no idea about, and the amount of scrutiny required is just too much. Because I have to assume you have no idea what you are doing.
π πairchild π
tante
James GleickDo you understand why Trump has lifted the oil sanctions that were punishing Iran, the country heβs bombing, and Russia, the country thatβs helping it?
Do you understand why Trump is sending 5,000 Marines to sail through the Strait of Hormuz while complaining that none of our friends and allies will help him reopen it?
Here's the whole explanation.
Eva WolfangelGanz frisch von mir. Die Wissenschaft hat festgestellt: das mit den KI-Agenten geht schief. (und zwar nicht, weil Superintelligenz und Weltuntergang, sondern weil Natur von KI-Systemen und so)
Freebie-Link fΓΌr euch
https://www.zeit.de/digital/datenschutz/2026-03/ki-agenten-studie-software-daten-sicherheit?freebie=b79c27e5
Freebie-Link fΓΌr euch
https://www.zeit.de/digital/datenschutz/2026-03/ki-agenten-studie-software-daten-sicherheit?freebie=b79c27e5
FrΓ©dΓ©ric JacobsβOur work demonstrates that πβππ‘ππ·ππ΄ is unreliable for the detection of illicit content: it is easy to incriminate someone by sending them false content with a hash value close to illicit content (a false positive) and to avoid detection of illicit content with minimal modifications to an image (a false negative)β
White-Box Attacks on PhotoDNA Perceptual Hash Function
πβππ‘ππ·ππ΄ is a widely deployed perceptual hash function used for the detection of illicit content such as Child Sexual Abuse Material (CSAM). This paper presents the first mathematical description of π΄ππππππ πβππ‘ππ·ππ΄, a new function which has identical outputs to that of πβππ‘ππ·ππ΄ for a large database of test images. From this description, several design weaknesses are identified: the algorithm is piece-wise linear and differentiable, the hash value only depends on the sum of the RGB values of each pixel, and it is trivial to find images with hash value equal to all zeroes. The paper further demonstrates that gradient-based optimization techniques and quadratic programming can exploit the mathematical weaknesses of π΄ππππππ πβππ‘ππ·ππ΄ and πβππ‘ππ·ππ΄ to produce visually appealing exact collisions and second preimages; for near-collisions and near-second-preimages the image quality can be further improved. The same techniques can be used to recover the rough shapes of an image from its hash value, disproving the claim from the designer that πβππ‘ππ·ππ΄ is irreversible. Finally, it is also shown that it is easy to produce high-quality perceptually identical images with a hash value that is far from the original image allowing to avoid detection. We have implemented our attacks on a large set of varied images and we have tested them on both π΄ππππππ πβππ‘ππ·ππ΄ and πβππ‘ππ·ππ΄. Our attacks have success rates close or equal to 100% and run in seconds or minutes on a personal laptop; they present a substantial improvement over earlier work that requires hours on parallel machines and that results only in near-collisions. We believe that with additional optimization of the parameters, the image quality and/or the attack performance can be further improved. Our work demonstrates that πβππ‘ππ·ππ΄ is unreliable for the detection of illicit content: it is easy to incriminate someone by sending them false content with a hash value close to illicit content (a false positive) and to avoid detection of illicit content with minimal modifications to an image (a false negative). False positives and leakage of information are particularly problematic in a Client Side Scanning (CSS) scenario as envisaged by several countries, where large hash databases would be stored on every user device and billions of images would be hashed with πβππ‘ππ·ππ΄ every day. Overall, our research cast serious doubts on the suitability of πβππ‘ππ·ππ΄for the large-scale detection of illicit content.
Ben Lockwood (broadly defined)
Jen Sorensen