4 ms·

Maybe I'm missing something, and this is a "dropbox" comment, but this article makes this sound like it's high-effort or high-skill. While I'm sure Neural Netw

Maybe I'm missing something, and this is a "dropbox" comment, but this article makes this sound like it's high-effort or high-skill. While I'm sure Neural Networks bring this to a new level, I feel like this wouldn't be hyper difficult to automate without them. What's more, $15 sounds like a lot. If this wasn't certainly going to be used for less-than-ethical purposes, I feel like it would be ripe for disrupting this market with $5 or $2 IDs of various quality.

nightpool 3y ago
I think what it comes down to is that most of the other ways to automate ID creation at scale were probably really easy to detect using ... well, probably neural networks. Because ID verification services had access to millions of "real" ID photos, I'd think a lot of fake images created using traditional automation would be pretty easy to detect (For example, having the same backgrounds/wood grain pattern reused for multiple images). So what neural network generation is adding here is basically being able to add much much more "plausible randomness" to the image (lens artifacts, dust motes, wood grain, plastic reflection, etc) in a way that makes distinguishing generated from non-generated images statistically very difficult.
- jszymborski 3y ago
  I'm not going to give these folks too many details, but there have long been ways to create convincing synthetic organic textures that predate NNs. Although I think you're right, the market tells us that this is likely either too high-skill and/or high-labour and/or easily detected.
- Muromec 3y ago
  > Because ID verification services had access to millions of "real" ID photos It's all snake oil devised to shift the risk from a party who has a mandate to perform KYC to some broken by design AI woodoo, which will generate free money until it doesn't once the landscape or the legislator's will mood change. They bank will then perform the actual KYC and will start asking questions once it sees something funny and will eat the cost of fraud at the end of the day.
jerf 3y ago
"this article makes this sound like it's high-effort or high-skill." It is. Today. Do not underestimate criminals. They have a full dark free market operating that you may be completely unaware of. The "Dark Web" is not just a media slogan. There is an entire economic ecosystem with high levels of specialization and structure already in it. There are already organizations smart enough to have people who can take this, productize it, and be the one selling the metaphorical shovels to the lower level criminals who buy this product and then take all the direct risks of doing the actual crime. Today's high-skill attack is tomorrow's product.
- jszymborski 3y ago
  I don't doubt criminals are highly skilled or motivated, nor that "onion" and "hidden" sites exist. I'm just not seeing why they would need to employ either great skill or motivation in this case. Just saying the task does not seem very difficult, but is described as being so. Would be happy to learn why it would be though!

nightpool 3y ago

I think what it comes down to is that most of the other ways to automate ID creation at scale were probably really easy to detect using ... well, probably neural networks. Because ID verification services had access to millions of "real" ID photos, I'd think a lot of fake images created using traditional automation would be pretty easy to detect (For example, having the same backgrounds/wood grain pattern reused for multiple images). So what neural network generation is adding here is basically being able to add much much more "plausible randomness" to the image (lens artifacts, dust motes, wood grain, plastic reflection, etc) in a way that makes distinguishing generated from non-generated images statistically very difficult.

jszymborski 3y ago
I'm not going to give these folks too many details, but there have long been ways to create convincing synthetic organic textures that predate NNs. Although I think you're right, the market tells us that this is likely either too high-skill and/or high-labour and/or easily detected.
Muromec 3y ago
> Because ID verification services had access to millions of "real" ID photos It's all snake oil devised to shift the risk from a party who has a mandate to perform KYC to some broken by design AI woodoo, which will generate free money until it doesn't once the landscape or the legislator's will mood change. They bank will then perform the actual KYC and will start asking questions once it sees something funny and will eat the cost of fraud at the end of the day.

jerf 3y ago

"this article makes this sound like it's high-effort or high-skill." It is. Today. Do not underestimate criminals. They have a full dark free market operating that you may be completely unaware of. The "Dark Web" is not just a media slogan. There is an entire economic ecosystem with high levels of specialization and structure already in it. There are already organizations smart enough to have people who can take this, productize it, and be the one selling the metaphorical shovels to the lower level criminals who buy this product and then take all the direct risks of doing the actual crime. Today's high-skill attack is tomorrow's product.

jszymborski 3y ago
I don't doubt criminals are highly skilled or motivated, nor that "onion" and "hidden" sites exist. I'm just not seeing why they would need to employ either great skill or motivation in this case. Just saying the task does not seem very difficult, but is described as being so. Would be happy to learn why it would be though!