Your subconscious can remember 30-character passwords

[Pythagorean]

your "subconscious" can remember 30-character passwords!

warning: not peer-reviewed

Apparently there's a new technique in password security. Only now, you don't have to worry about forgetting your password because it's not stored in autobiographical memory (the memory you're explicitly conscious of) but instead stored in procedural memory (i.e. basal ganglia, pavlov's dog, etc).

http://www.extremetech.com/extreme/...r-password-in-your-brains-subconscious-memory

 

[Evo]

If you don't consciously know your password, how can you consciously use it? I saw no explanation. "Seemingly", "expected to perform better"?

 

[Pythagorean]

you play a game:

Authentication requires that you play a round of the game — but this time, your 30-letter sequence is interspersed with other random 30-letter sequences. To pass authentication, you must reliably perform better on your sequence. Even after two weeks, it seems you are still able to recall this sequence.

Your micro-actions (how well you perform on each keystroke) during the grame will be influenced by how you were trained on the game, which will be specific to your password. You have to perform better on your sequence (meaning your timing is better because you're brain is more-expecting the next letter to be congruent with your sequence) than you do on other (random) sequences within a single game session

example:

if a subsequence in your password is "J-K-L-J-M-N' then as you play the game, you'll hit that packet and, because you've trained with it, you'll always be expecting a K or an M after the J so you'll have a better response time . When it ends up being an N after a J, and you're expecting a K or an M, switching to the N will slow you down, further accentuating the difference in timing performance between stroke series outside your sequence vs. inside your sequence.

 

[Evo]

you play a game:



Your micro-actions (how well you perform on each keystroke) during the grame will be influenced by how you were trained on the game, which will be specific to your password. You have to perform better on your sequence (meaning your timing is better because you're brain is more-expecting the next letter to be congruent with your sequence) than you do on other (random) sequences within a single game session

But did anyone actually manage to reproduce the exact 30 character password they were given, much less reproduce it repeatedly? It sounds like the answer is no.

 

[Pythagorean]

The article claims that yes, up to two weeks later you can still recall it:

By this point, their experimental results suggest that the 30-letter password is firmly implanted in your subconscious brain. Authentication requires that you play a round of the game — but this time, your 30-letter sequence is interspersed with other random 30-letter sequences. To pass authentication, you must reliably perform better on your sequence. Even after two weeks, it seems you are still able to recall this sequence.

(they include a plot of the data)

Which is why it's going to be presented at the Usenix Security Symposium this August.

 

[Pythagorean]

Here's a more thorough report from the authors of the system (still don't think it's peer-reviewed):

http://bojinov.org/professional/usenixsec2012-rubberhose.pdf

 

[Evo]

So, this is customized to each individual's abilities and based on their improvement on sequences at one and two week intervals.

 

[Pythagorean]

I wouldn't say improvement so much as retention. They are not trained for the two weeks, they just return to take the test. Though it's somewhat confounding that they double the test size for the two week interval. You can't change two variables at once!

But yeah, it looks like only about half the people were able to retain the sequence.

 

[atyy]

Hmmm, is this related to Rapid formation of robust auditory memories: insights from noise?

"Here we used random waveforms to probe the formation of new memories for arbitrary complex sounds. A behavioral measure was designed, based on the detection of repetitions embedded in noises up to 4 s long. Unbeknownst to listeners, some noise samples reoccurred randomly throughout an experimental block. Results showed that repeated exposure induced learning for otherwise totally unpredictable and meaningless sounds. The learning was unsupervised and resilient to interference from other task-relevant noises. When memories were formed, they emerged rapidly, performance became abruptly near-perfect, and multiple noises were remembered for several weeks."

 

[Pythagorean]

I'd assume that in both cases, the end target for procedural ("subconscious") memory storage is the basal ganglia, but it may be different regions of the basal ganglia with different encoding rules.

I don't know what the functional differences are between motor/visual and auditory. There's probably bound to be some important differences, seeing as how we tend to emphasize visual stimuli as humans.

 

[atyy]

I'd assume that in both cases, the end target for procedural ("subconscious") memory storage is the basal ganglia, but it may be different regions of the basal ganglia with different encoding rules.

Would http://www.ncbi.nlm.nih.gov/pubmed/20519543 be evidence against this hypothesis?

 

[Pythagorean]

Hrm... they only seem to look at the globus pallidus.. that's not the whole basal ganglia and that's specifically the motor output of the basal ganglia...

The striatum (the other "half" of the basal ganglia) would be where sensorimotor information is received :)

So, to simplify, they're looking at the output and saying "hey there's only output!". Maybe I'm missing something, but that's not an impressive discovery to me. I also have no idea about the homology of the basal ganglia between humans and primates.

 

[atyy]

Hrm... they only seem to look at the globus pallidus.. that's not the whole basal ganglia and that's specifically the motor output of the basal ganglia...

The striatum (the other "half" of the basal ganglia) would be where sensorimotor information is received :)

So, to simplify, they're looking at the output and saying "hey there's only output!". Maybe I'm missing something, but that's not an impressive discovery to me. I also have no idea about the homology of the basal ganglia between humans and primates.

But isn't the striatum usually considered "upstream" of the globus pallidus? So if the motor action is downstream of GP, and GP is downstream of the striatum, then shouldn't the striatum be effectively inactivated too?

 

[Pythagorean]

I'm not sure of you're argument. You mean to imply that the only way to output from the striatum is through the GP? But that is not the case and homology becomes important here: one of the important alternative pathways (from the striatum to the supplementary motor area) leads to an area that is much more pronounced in nonhuman primates than in humans (having a recognizable body mapping in primates). But striatum also has pathways to the substantia nigra and superior colliculus (also involved in motor function/planning). So is all pathways from the the striatum to some motor activity cut off? I don't see why.

The difficult thing about this question is that there is redundancies all over the system. So to what extent does each part plays a role in some event we call a "motor task" is undoubtedly a complicated question, and to what extent functionality overlaps is an even more difficult question.

The paper you linked provides references to the evidence for striatum involvement in procedural learning in their introduction (and of course, challenges it with conjecture in the discussion section).

Not to say that my hypothesis correct and the final storage area is the BG, but I don't think the methodology in that paper really confronts the issue any more than superficially.

 

[atyy]

I'm not sure of you're argument. You mean to imply that the only way to output from the striatum is through the GP? But that is not the case and homology becomes important here: one of the important alternative pathways (from the striatum to the supplementary motor area) leads to an area that is much more pronounced in nonhuman primates than in humans (having a recognizable body mapping in primates). But striatum also has pathways to the substantia nigra and superior colliculus (also involved in motor function/planning). So is all pathways from the the striatum to some motor activity cut off? I don't see why.

The difficult thing about this question is that there is redundancies all over the system. So to what extent does each part plays a role in some event we call a "motor task" is undoubtedly a complicated question, and to what extent functionality overlaps is an even more difficult question.

The paper you linked provides references to the evidence for striatum involvement in procedural learning in their introduction (and of course, challenges it with conjecture in the discussion section).

Not to say that my hypothesis correct and the final storage area is the BG, but I don't think the methodology in that paper really confronts the issue any more than superficially.

Yes, it could be that the "habit" function of the striatum is outputted so robustly through GP, that their GP inactivation was insufficient, or that it does not go through GP. I would be a bit surprised by the latter, but of course it's not impossible, since my prior is based mainly on anatomy. But do you think there are loopholes other than those (eg. in experiment design?)

 

[Pythagorean]

The only other thing that raised a question in my mind was that the training period was six months. What if this gives the organism as chance to embed the memories at a more explicit level?

Also, any time we assay a memory task, because there are several layers of memory involved... what if detailed procedural memories are actually being lost, but they can be remodeled coarsely with other neural systems using a more coarse-grain memory from another part of the brain? The result would look like a performance loss.

Here's where I learned of hippocampal vs. basal ganglia memory roles, btw:

Based on prior animal and computational models, we propose a double dissociation between the associative learning deficits observed in patients with medial temporal (hippocampal) damage versus patients with Parkinson's disease (basal ganglia dysfunction). Specifically, we expect that basal ganglia dysfunction may result in slowed learning, while individuals with hippocampal damage may learn at normal speed. However, when challenged with a transfer task where previously learned information is presented in novel recombinations, we expect that hippocampal damage will impair generalization but basal ganglia dysfunction will not. We tested this prediction in a group of healthy elderly with mild-to-moderate hippocampal atrophy, a group of patients with mild Parkinson's disease, and healthy controls, using an "acquired equivalence" associative learning task. As predicted, Parkinson's patients were slower on the initial learning but then transferred well, while the hippocampal atrophy group showed the opposite pattern: good initial learning with impaired transfer. To our knowledge, this is the first time that a single task has been used to demonstrate a double dissociation between the associative learning impairments caused by hippocampal versus basal ganglia damage/dysfunction. This finding has implications for understanding the distinct contributions of the medial temporal lobe and basal ganglia to learning and memory.

Dissociating Hippocampal versus Basal Ganglia Contributions to Learning and Transfer
http://dl.acm.org/citation.cfm?id=1162419

 

[atyy]

The only other thing that raised a question in my mind was that the training period was six months. What if this gives the organism as chance to embed the memories at a more explicit level?

Also, any time we assay a memory task, because there are several layers of memory involved... what if detailed procedural memories are actually being lost, but they can be remodeled coarsely with other neural systems using a more coarse-grain memory from another part of the brain? The result would look like a performance loss.

That's an interesting thought.

Here's where I learned of hippocampal vs. basal ganglia memory roles, btw:Dissociating Hippocampal versus Basal Ganglia Contributions to Learning and Transfer
http://dl.acm.org/citation.cfm?id=1162419

In a review Turner and Desmurget propose the basal ganglia are involved in acquisition, but not long term storage - which sounds to me like basal ganglia are to procedural memories what the hippocampi are for declarative ones. So maybe there isn't any conflict with what Myers et al are saying?

 

[RabbitWho]

Why couldn't I just make up a 30 character sentence with easy to remember spelling/grammatical mistakes?

E.g. I probally buy milk from cow's all be it indirectly !11

 

[Pythagorean]

That's an interesting thought.



In a review Turner and Desmurget propose the basal ganglia are involved in acquisition, but not long term storage - which sounds to me like basal ganglia are to procedural memories what the hippocampi are for declarative ones. So maybe there isn't any conflict with what Myers et al are saying?

Another thread I responded to reminded me of this thread:

https://www.physicsforums.com/showthread.php?t=657465

Now that I'm done with my thesis, I have an opportunity to look at some of these papers. The review casts doubt on the basal ganglia as storage and suggests the motor cortex. In the context of the thread above, I wonder what that means for reptiles, who have a very tiny homolog to the motor cortex. Or I guess, more importantly, what it means from humans. Does our greater complexity come from simply having larger (and presumably more specialized) memory banks, and not so much the way we process information on the fly?

But this also seems to be exclusive to motor tasks. What about, say, "procedural thinking" tasks?