Epsilon: Exam mode can be (I think) bypassed easily

[Zezombye]

As I said here, all it takes for the exam mode to be bypassed is to remove the lines of code that block memory access, recompile the OS, then transfer it on the calculator.

Now, I could be wrong and one couldn't do that (I can't test since I don't have access to the real calculator), but if it is possible you should prevent this.

[boricj]

I've already nuked the exam mode on the simulator with a single line comment. There's no reason why it wouldn't work on hardware.

The HP Prime hardware does not offer any root of trust capability and has no flashing restriction. Heck, I've demonstrated bare metal LED blinking in 2015, a dumb third-party RPN calculator firmware in 2016 and somebody ported the Linux kernel this week. The only practical difference in killing its exam mode when compared to NumWorks would be the amount of reverse engineering to perform on its closed-source firmware.

The HP Prime was introduced in 2013. I haven't heard the French Éducation Nationale freak out about a completely unlocked calculator from frakking _HP_.

The whole point of this project is empowering calc users with freedom to modify their calculator firmware and run whatever firmware they want. Take it away, I'm out.

I'll spare you the argument why exam mode is a dumb idea. But I'll just say this: you can download the firmware of a NumWorks calculator in DFU mode, compare it byte-for-byte with a reproductible build of its source code and prove you did not tamper with exam mode even with a homemade customized firmware. Try to find another calculator where you can independently audit its exam mode and not just blindly trust the manufacturer.

[millerjs2]

There needs to be a way for an exam mode to persist alongside the ability to have a totally open device.

The fact of the matter is, for NumWorks to succeed they need to sell at a volume higher than hobbyists and enthusiasts alone can provide. High school and university students are by far and above the largest purchasers of calculators, and whether or not they can use those calculators on standardized tests is a major factor.

[debrouxl]

The HP Prime, TI-84+CE / TI-83PCE, and at least the two Casio models sold in France as Graph 35+E and 75+E aren't more secure than the NumWorks calculator. On all of those, the exam mode can be easily tampered with - the manufacturers' only protection is that the ones with a technical ability to tamper with the exam modes hold themselves to moral standards.

There's no reasonable way to ban the NumWorks calculator while leaving the more usual commercial models from big, well-established manufacturers untouched. In fact, banning the open NumWorks calculator but not models from TI, Casio and HP would be a PR disaster for whatever standardized testing regulation authority dares making such a blunder... and TI/Casio/HP would quickly resent such a decision too, given that an instance of the well-known Streisand effect would attract much attention to their weak platforms, and if the right persons work on the matter, all platforms would fall in a matter of hours (at worst, days).

We've seen retaliation at play on the Nspire series, after TI decided to block unofficial access to native code (through the Ndless framework) too many times, despite the crystal clear usefulness of native code. "PTT killers" in a matters of days (censored on the main forums of the community but available elsewhere), later the ability to run arbitrary OS seamlessly through exploiting the boot2's holes, etc.

[Zezombye]

On all of those, the exam mode can be easily tampered with - the manufacturers' only protection is that the ones with a technical ability to tamper with the exam modes hold themselves to moral standards.

That's a crucial difference IMO - it means that most numworks calculators will have a tampered exam mode (because you only need to comment a few lines then recompile the OS), whereas few casio/TI calculators will be tampered with, simply because it's way harder to disassemble code and figure out how to remove the memory restriction.

If the government sees that most cases of cheating are on numworks, they could ban them. Sure, the community could retaliate by releasing tweaked OSes in the wild, and students could cheat using any calculator, but they could outright ban calculators like they've probably planned to do :/

Maybe I'm fear-mongering too much, but it would suck to have this calculator banned.

[adriweb]

Let’s be real here: for any tempering programs, end users would simply execute/install an OS/program they’d have downloaded somewhere. No source code modification involved.
So it doesn’t matter in this regard if the NumWorks OS is open source in the first place. Not only does that not matter at all for people using cheating tools, but as I was saying, end users don’t even know (nor care) about it: they simply launch something already built for them.
So all the calc models listed above are on the same boat. And the government seems not to care, at least for now (don’t forget most communities ban such programs anyway, so...).
In addition, as @boricj was saying, the NumWorks might actually be the only platform where it’s technically possible to check the state of things manually without having to rely on trusting a third party.

Anyway, in conclusion, I wouldn't really be worried - but as developers and/or enthusiasts, we might as well not add virtual fuel to the potential fire that some governement people might see/imagine (so let's continue not to promote any cheating tools/patchs/forks etc., it'd hurt everyone) :)

[Ecco]

So all the calc models listed above are on the same boat (...). as developers and/or enthusiasts, we might as well not add virtual fuel to the potential fire

I think this is a good conclusion.

I'm going to close the issue because no change is going to be made to the firmware on this matter, in one direction or the other.