[EFats]

Why are you concluding that the odds are one in ten for getting it? ,,,

Becase it is when you only use the 10 possible digits. The slightly technical explanation:

Suppose the actual number is at grid position [x,y] (and I mean the number grid, not the picture location, which is irrelevant in this attack). I can choose any number I want to look at, say "8" which is at some other grid position [j,k]. I have no idea what [x,y] or [j,k] are, but that actually does not matter. Once I have picked my number "8" at grid [j,k], then all I need to know is that the number at [x,y] can only be one of 10 possibilities (0...9). Therefore, IF (and that can be a big assumption), I know which grid size is used, every random shuffle of the number grid has a 1/10 = 10% possibility of having your special number at grid [x,y]. For 5 tries, that means you have 50% chance of hitting the right number. Those are pretty good odds.

As someone pointed out, you may not need to be that observant if there are only 2 grid sizes (either sparse or dense) used on the Z10 or Z30. If that's the case, it would be really easy for someone who knows how the BB10 picture password system works to have a go at unlocking the phone.
That's why allowing ANY character can boost up the security significantly, you can go from 10% to 1/36 if you include alphanumeric and if you include punctuation and other symbols it can go up from there. The easiest fix is simply to have random number placement, not even a regular grid. Well, actually allowing alphanumerics is pretty easy too but it doesn't buy you as much as a random number placement.

The standard BB10 password lock should actually work quite well since it can be (almost) arbitrary length of characters and with BB10, the press & hold entry method for upper case is an additional obfuscation to what character you are actually entering

[spARTacus]

Becase it is when you only use the 10 possible digits. The slightly technical explanation:.....

Yeah I got it from the other logical explanation and from re-reading the other earlier posts, which I admit I must not have read fully first time through. Good additional details however. Thanks. ☺

Posted via CB App for Android on Tab4 (interim Playbook replacement)

[Richard Buckley]

Actually they kind of have.
There were some links posted above.
They even used the words shoulder surfing. Admittedly I don't think they ever said it was perfect, opting for words like "hard" instead, but given how small a change it would take to make it way harder, I think they should have done it.

You may be referring to this:

Shoulder surfing. With no set points to track, only a number and a location that can be obtained from anywhere on the screen, prying eyes will find it very hard to figure out how I access my phone.

And this is true. If I watch someone type a password on a keyboard it is a direct process of noting what keys are pressed and in what order. Picture Password is more difficult (I wouldn't use the adjective very) as evidenced by the original post, and that Kevin thought it would be hard for people to guess his Picture Password from a video. You can call it a vulnerability if you want, but it is by no means one that is limited to Picture Password over other password entry methods, and it does provide protection that other methods don't, and is resistant to other more usable methods of determining passwords such as the paper that I also linked. These are all properties that would tend to cause security researchers to not call it a vulnerability unless they were trying to excite people.

LeapSTR100-2/10.3.3.2205

[1122334455667788]

You may be referring to this:

And this is true. If I watch someone type a password on a keyboard it is a direct process of noting what keys are pressed and in what order. Picture Password is more difficult (I wouldn't use the adjective very) as evidenced by the original post, and that Kevin thought it would be hard for people to guess his Picture Password from a video. You can call it a vulnerability if you want, but it is by no means one that is limited to Picture Password over other password entry methods, and it does provide protection that other methods don't, and is resistant to other more usable methods of determining passwords such as the paper that I also linked. These are all properties that would tend to cause security researchers to not call it a vulnerability unless they were trying to excite people.

LeapSTR100-2/10.3.3.2205

I don't completely disagree.
I also don't share the outrage that some seem to feel at the admittedly rather dramatic title.
The fact is it is more vulnerable to shoulder surfing that many people (myself included) had believed. This may not be enough to justify a security label of Vulnerable, but it's also not a completely incorrect usage of the word.


I'll also point out that when described in a blog post in 2014 they said:
"Shoulder Surf Attack – When using a simple password, someone could look over your shoulder while you enter it. Picture Password prevents this situation from occurring by drawing a random number grid and varying the size of the grid. For example, in addition to a random number grid, the grid size also randomly changes increasing and decreasing the amount of rows and columns to reduce the shoulder attack vulnerability."
Notice the word prevents? That's sufficient that I would expect better than 50% odds even against someone with great memory.
They also refer to the shoulder attack as a vulnerability .

If I remember Kevin's contest correctly it wasn't actually a valid test of the system. I think you had to figure out the specific number and location (both things you never actually figure out using this attack method).

[EFats]

The spacing is reused. You can simply count the number of columns and rows.

As far as I can tell it is reused, but on my Z10, the column varies from 5 to 8, I haven't checked the row count but this would give at least 4 different grid sizes an attacker would have to note.

[1122334455667788]

As far as I can tell it is reused, but on my Z10, the column varies from 5 to 8, I haven't checked the row count but this would give at least 4 different grid sizes an attacker would have to note.

I don't see why it really matters. Anyone who has good enough memory (or a photo/video) to pick a number and remember it's exact location is likely able to remember the number of columns/rows too.
They don't need to keep track of how many options there are. They'll be ignoring all the non-matching ones anyways.

[misterabrasive]

The spacing is reused since it only consists of whole rows and columns, and seems to be selected from a relatively small range, between 6 and 8 columns on my Leap. I have not been able to determine any procedure for selection so it is probably random, but proving that would take more time than I'm will to invest.

The bottom line is that if you allow someone to watch you enter your secret credentials, regardless of the method of entry, you are reducing the strength of that protection. This can only be a vulnerability if BlackBerry claims that Picture Password is robust against shoulder surfing. To my knowledge they have never made that claim. This is why there are "protect your PIN" on bank machines etc.

So far however Picture Password is robust against spatiotemporal dynamics, which is enough for me.

LeapSTR100-2/10.3.3.2205

Yes, and the bottom line to this thread is that it was the OP and his 'friend' who probably spend a lot of time looking over each other's shoulders. If you are lazy enough to unlock your phone in a crowd and you then have your phone stolen by the person standing behind you, well..... There's probably no security measures that will overcome stupidity.

[Thud Hardsmack]

Becase it is when you only use the 10 possible digits. The slightly technical explanation:

Suppose the actual number is at grid position [x,y] (and I mean the number grid, not the picture location, which is irrelevant in this attack). I can choose any number I want to look at, say "8" which is at some other grid position [j,k]. I have no idea what [x,y] or [j,k] are, but that actually does not matter. Once I have picked my number "8" at grid [j,k], then all I need to know is that the number at [x,y] can only be one of 10 possibilities (0...9). Therefore, IF (and that can be a big assumption), I know which grid size is used, every random shuffle of the number grid has a 1/10 = 10% possibility of having your special number at grid [x,y]. For 5 tries, that means you have 50% chance of hitting the right number. Those are pretty good odds.

As someone pointed out, you may not need to be that observant if there are only 2 grid sizes (either sparse or dense) used on the Z10 or Z30. If that's the case, it would be really easy for someone who knows how the BB10 picture password system works to have a go at unlocking the phone.
That's why allowing ANY character can boost up the security significantly, you can go from 10% to 1/36 if you include alphanumeric and if you include punctuation and other symbols it can go up from there. The easiest fix is simply to have random number placement, not even a regular grid. Well, actually allowing alphanumerics is pretty easy too but it doesn't buy you as much as a random number placement.

The standard BB10 password lock should actually work quite well since it can be (almost) arbitrary length of characters and with BB10, the press & hold entry method for upper case is an additional obfuscation to what character you are actually entering

Wasn't until I read this that I understand what the OP is getting at, and I still think it's extremely unlikely. I've noticed the path the numbers have to take varies greatly, it's never a straight line shot and sometimes the distance is such that a second try is required due to lack of real estate for a thumb or finger to move the correct number to the unlock zone. I'm usually drawing a curved path and overshooting the target because of the animation acceleration, which then means another smaller curved path back. I still can't see that moving numbers randomly can have that high of a success rate. Excellent explanation by the way.

[Thud Hardsmack]

As far as I can tell it is reused, but on my Z10, the column varies from 5 to 8, I haven't checked the row count but this would give at least 4 different grid sizes an attacker would have to note.

Z10 has 3: 10x8, 9x7, and 8x6.

[bathu]

What will happen after 10 wrong password attempts in picture password?

Posted via CB10

[Thud Hardsmack]

What will happen after 10 wrong password attempts in picture password?

Posted via CB10

Security wipe.

[bathu]

Is it possible to increase the password attempts like android powered blackberry device?

Posted via CB10

[Thud Hardsmack]

Is it possible to increase the password attempts like android powered blackberry device?

Posted via CB10

Not possible.

[Richard Buckley]

...
"Shoulder Surf Attack – When using a simple password, someone could look over your shoulder while you enter it. Picture Password prevents this situation from occurring...

Saying that it prevents someone seeing your simple (or complex if they are an accomplished shoulder surfer) password while looking over your shoulder, because you aren't entering a password.

If I remember Kevin's contest correctly it wasn't actually a valid test of the system. I think you had to figure out the specific number and location (both things you never actually figure out using this attack method).

It was a valid test of the system, just not the same as this test because if he had typed his password on the video it would have been a race to see who could post first, and with video recording it would not have to be a simple password either. Also if I remember correctly someone did correctly guess the number and position. This is also a simple matter of probability there are on 10 choices of number. If you can do some social engineering on the target you should be able to reduce the number of candidate locations for their picture. If not there are still a limited number of valid locations determined by the allowed error circle. If you have routine access to a phone you could try all combinations a few at a time. Ultimately the amount of protection comes down to the amount of entropy available in the system. The amount of entropy in a Picture Password is on the same order of magnitude as a four digit PIN or swipe pattern. It is not anywhere nearly as vulnerable to shoulder surfing as those methods.

If there are people using picture password who believe that it removes all risks of letting someone watch while they unlock their phone then this thread serves a good service of informing them that is not the case. But like any IT security issue, if you over hype it a lot of people will just give up because they feel that being secure is too hard. I see that all the time. That's why I recommend using as long and as complex device password as you can remember, set the lock timer as short as you can live with, but use picture passport (or swipe on or whatever) to unlock the phone. Then when you suspect someone is shoulder surfing or trying to exhaust over all possible number location combinations you can immediately go somewhere private and change your picture password. You don't have to change your good password because it wasn't compromised. A new picture password is easy to remember, a new good password isn't.

LeapSTR100-2/10.3.3.2205

[1122334455667788]

What will happen after 10 wrong password attempts in picture password?

Posted via CB10

Actually you don't get 10 tries.
You get 5, and then it switches to the regular password.
The only exception to this being the Priv with Lollipop 🍭.

[Dmd74]

I have to be honest, this seems like a stretch. I would be shocked if someone could unlock my phone using the Picture Password.

Just another comment, I feel as though the picture password is greatly underappreciated This way to unlock the phone is amazingly simple but extremely effective.

[Richard Buckley]

So there has been some discussion of the odds here and there without looking at the actual math. So lets make some reasonable assumptions as see where that leads us.

First assumption. Varying grid pitch could delay an attacker, but if the attacker does not make any guesses at the pitch sizes not observed they don't count against the allowed incorrect guess count. So we must assume that the attacker has enough time to continuously reject bad pitch attempts. (This suggests a change to Picture Password that would improve its resistance but result in less convenience. Always a trade off with security features.)

Second assumption, the attacker is not concerned with leaving the phone at the point of "Enter BlackBerry to continue" after using five attempts.

As other people have stated, once the variable pitch size is eliminated the problem one of the relationship of the correct number with respect to the attacker's chosen number. Or what is the probability that the randomly selected number X columns left or right and Y rows up or down (where the values of X, Y, left or right and up or down are unknown but finite) is the number chosen by the user. Assuming the grid is truly randomly populated the the odds are 1 in 10. So what are the odds that the attacker will break into the phone in the five tries available? This is actually very similar to the problem posed here.

As with dice the math is greatly simplified if we consider the odds of the correct number not being in the correct spot each time, similar to considering the odds of not rolling a six on each dice. The odds of the number being wrong are 9 in 10. So the odds of the attack failing purely due to chance in five attempts are 9/10 x 9/10 x 9/10 x 9/10 x 9/10 = 59,049/10,000 or 59%. The odds of the attacker succeeding are 40,951/100,000 or 41%.

So in order to determine if Picture Password is more or less vulnerable to shoulder serfing than other methods we must decide if an attacker who can pick a number and location and identify the correct pitch size with suffecient accuracy is able to read your PIN, password, swipe pattern etc under the same conditions at more or less than 41% probability.

[JuiciPatties]

Security wipe.

Actually you don't get 10 tries.
You get 5, and then it switches to the regular password.
The only exception to this being the Priv with Lollipop ������.

That is correct, you only get 5 tries with picture password. This is an important aspect of the security built into it because it could be a kid playing with the phone or pocket entries and you don't want to cause an accidental security wipe if you allow it to continue. After the 5th try, you need to type in "blackberry" before you can try the normal password. The chances of an accidental or pocket entry of "blackberry" is very, very slim. Efats or someone else in this thread can calculate that probability, and if it is >1%, I would be shocked!!

[Thud Hardsmack]

That is correct, you only get 5 tries with picture password. This is an important aspect of the security built into it because it could be a kid playing with the phone or pocket entries and you don't want to cause an accidental security wipe if you allow it to continue. After the 5th try, you need to type in "blackberry" before you can try the normal password. The chances of an accidental or pocket entry of "blackberry" is very, very slim. Efats or someone else in this thread can calculate that probability, and if it is >1%, I would be shocked!!

That's correct, I was only thinking the overall total attempts. Thanks for making sure that's clarified!

[Thud Hardsmack]

So there has been some discussion of the odds here and there without looking at the actual math. So lets make some reasonable assumptions as see where that leads us.

First assumption. Varying grid pitch could delay an attacker, but if the attacker does not make any guesses at the pitch sizes not observed they don't count against the allowed incorrect guess count. So we must assume that the attacker has enough time to continuously reject bad pitch attempts. (This suggests a change to Picture Password that would improve its resistance but result in less convenience. Always a trade off with security features.)

Second assumption, the attacker is not concerned with leaving the phone at the point of "Enter BlackBerry to continue" after using five attempts.

As other people have stated, once the variable pitch size is eliminated the problem one of the relationship of the correct number with respect to the attacker's chosen number. Or what is the probability that the randomly selected number X columns left or right and Y rows up or down (where the values of X, Y, left or right and up or down are unknown but finite) is the number chosen by the user. Assuming the grid is truly randomly populated the the odds are 1 in 10. So what are the odds that the attacker will break into the phone in the five tries available? This is actually very similar to the problem posed here.

As with dice the math is greatly simplified if we consider the odds of the correct number not being in the correct spot each time, similar to considering the odds of not rolling a six on each dice. The odds of the number being wrong are 9 in 10. So the odds of the attack failing purely due to chance in five attempts are 9/10 x 9/10 x 9/10 x 9/10 x 9/10 = 59,049/10,000 or 59%. The odds of the attacker succeeding are 40,951/100,000 or 41%.

So in order to determine if Picture Password is more or less vulnerable to shoulder serfing than other methods we must decide if an attacker who can pick a number and location and identify the correct pitch size with suffecient accuracy is able to read your PIN, password, swipe pattern etc under the same conditions at more or less than 41% probability.

I still don't see it happening. Here's the small (8x6) grid on my Z10 with the zone marked and all correct numbers highlighted - how is it possible to guess in less than five tries to get any of those to that location without prior knowledge if I'm not placing my thumb directly on the number? The grid size and number layout are never repeated.

[1122334455667788]

As with dice the math is greatly simplified if we consider the odds of the correct number not being in the correct spot each time, similar to considering the odds of not rolling a six on each dice. The odds of the number being wrong are 9 in 10. So the odds of the attack failing purely due to chance in five attempts are 9/10 x 9/10 x 9/10 x 9/10 x 9/10 = 59,049/10,000 or 59%. The odds of the attacker succeeding are 40,951/100,000 or 41%.

I stand corrected.
I kind of thought I was missing something with the math.

[1122334455667788]

I still don't see it happening. Here's the small (8x6) grid on my Z10 with the zone marked and all correct numbers highlighted - how is it possible to guess in less than five tries to get any of those to that location without prior knowledge if I'm not placing my thumb directly on the number?

You haven't lined up the grid properly.
The point is that although you don't know the correct number, you can know a correct grid alignment

[Thud Hardsmack]

You haven't lined up the grid properly.
The point is that although you don't know the correct number, you can know a correct grid alignment

That's how the grid is placed by the device.

[1122334455667788]

That's how the grid is placed by the device.

You have to move it.

[Thud Hardsmack]

You have to move it.

So how would an attacker be able to determine where to move it to line it up?