NSA provides list of patents it's willing to share with public
Nevermind that the NSA currently is paying telecommunication companies to store our metadata (thanks Snowden!) or that the NSA is subsidizing transoceanic cables (non sarcastic thanks Snowden) https://www.nsa.gov/what-we-do/research/technology-transfer/assets/files/nsa... Patent 8,363,825, from 2009, looks at lot like some kind of CRC fuzzy hashing. Patent 9,525,866, from 2016, is anti-camera forensics. Time to go back to film! (or maybe just resize it to 50% and save it as a rather lossy JPEG) Patent 7,406,595, from 2004, shows how the NSA encrypts communications. Sounds more efficient than Enchilada, as submitted to CAESAR. Patent 6,922,774, from 2001, shows what the NSA does with virtual machines. Patent 6,724,893, from 1996, is about key escrow. If one uses TENS, a version of linux from the US government, one notices that the help manual for the encryption wizard talks about key escow. But all our keys is escrowed since the NSA stole SIM card keys from foreign manufacturers of our devices? Patent 6,912,284, from 1983 and the longest expiration date, is about some kind of LFSR AEAD cipher. What? They discovered this in 1983 and we're still working on AEAD ciphers? Patent 6,820,830 appears to be just placing slants under a paper shredder to direct output into multiple rubbish bins. Like people. You can just go to the NSA's website. This is two clicks away from the front page. What are you people doing? If you're going to be lazy.
On 07/07/2017 04:17 PM, Ryan Carboni wrote:
Patent 9,525,866, from 2016, is anti-camera forensics. Time to go back to film! (or maybe just resize it to 50% and save it as a rather lossy JPEG)
I would not trust that very far: Signatures from hot pixels may remain visible to the naked eye as well as statistical analysis, and variations in color registration across larger areas of the sensor would survive. A camera signature does not need to be unique to be very valuable to an adversary who has other means at hand to narrow the field of candidates for creator of this or that image of interest. Conversion to a vector format with the aid of an auto-trace program should work reliably, at the cost of a /massive/ loss of resolution. If the photo needs to still look like a photo when you're done sanitizing it, the GMIC filter pack's tool set may be useful, with effects like bilateral smoothing etc. that wipe out small scale noise signatures. Conversion to grayscale, or or reducing color resolution and changing RGB balance would also be a good idea. Exporting sanitized images in an indexed format with a limited size palette (i.e. GIF) would be helpful. But no matter what, if you remove enough detail to confidently disassociate the photo from the camera, image quality will suffer quite visibly. One's method for altering photos to disassociate them from individual camera sensors would have to change often or it would create a trail of its own. :o)
On 07/08/2017 11:23 AM, Steve Kinney wrote:
On 07/07/2017 04:17 PM, Ryan Carboni wrote:
Patent 9,525,866, from 2016, is anti-camera forensics. Time to go back to film! (or maybe just resize it to 50% and save it as a rather lossy JPEG)
I would not trust that very far: Signatures from hot pixels may remain visible to the naked eye as well as statistical analysis, and variations in color registration across larger areas of the sensor would survive.
[snip] Postscript: Or pay cash for a cheap toy camera, destroy the sensor and discard after use. Duh. Not always an option depending the situation, but probably the best one where practical. :o)
participants (2)
-
Ryan Carboni -
Steve Kinney