Sanity Check: Cyborg and AI Uploads

[Surrealistik]

These make antaging as Sci way too easy: build an upload console virtually anywhere, stick in your pre-fab airtight laws, win. This can all be done without arousing suspicion so long as you choose an opportune time to mint the upload and freeform circuitry (which should be pretty easy between Sci being a pretty boring place and privacy shutters).

Two fixes:

A: You can only link and upload to Cyborgs and AIs on the same Z level; no more building uploads on the Research Outpost, for almost guaranteed undetectability (which to be fair still exists on the station Z level anyways).

and

B: The Captain and/or Head of Security IDs are required to authenticate an Upload in order to link it to existing AIs and Cyborgs via physical insertion/swiping. Once an upload is authenticated, the IDs are no longer necessary until it is deauthenticated at the Upload. The Uploads in the AI core start off authenticated.

[Lohikar]

Why would the Head of Security be able to authorize an upload? The AI is under the CE & RD's authority.

[icy_dew]

a different law would be better to be handled by the HoS than the RD, because otherwise the RD can decide by himself wether a law is suitable for the integrity of the station. the HoS needs to go over the security of the employees.

[Guest Complete Garbage]

Why not just require the ID of any two heads of staff or the Captain? Pretty simple, considering current directives support it, and also bypassable via a single EMAG swipe.

Edit: At code green, it should either:

1. Only require one head of staff's ID

2. Require one head's ID and one roboticist's ID, OR only the Captain's ID.

At code blue or above:

1. 2 heads of staff

2. 1 head AND Captain's

3. 2 heads OR Captain's

This would fit better with directives.

Reset boards shouldn't be locked, and should be implementable by anyone with access to a console, simply as a safety precaution.

[Surrealistik]

If you're the sole antag, it's pretty easy for Green to still be in place by the time you get the items required for subversion; I'm pretty leery about requiring only one Head for that reason. Even though this change might help with all of Research being OP as antag the RD is still insanely advantaged, even as heads go.

Beyond that, merely building Uploads should probably be illegal without the authorization of two heads unless it's an emergency.

[Lohikar]

If you're the sole antag, it's pretty easy for Green to still be in place by the time you get the items required for subversion; I'm pretty leery about requiring only one Head for that reason. Even though this change might help with all of Research being OP as antag the RD is still insanely advantaged, even as heads go.

Beyond that, merely building Uploads should probably be illegal without the authorization of two heads unless it's an emergency.

I'm fairly certain that it is already against regs to build uploads without authorization.

[Fiskap]

The station directives probably allow you to build that within science, for research purposes.

[Chada1]

Surrealistik, there are Station Directives. Specifically, Directive 7. Regarding AI upload access - Station Directive 7

 

To clarify who, when and under what terms personnel may enter the AI upload.

In standard operation access to the upload requires one head of staff present with, approval of other members of the command staff. In non-standard operation access to the upload requires two present heads of staff. Approval of the other members of command is recommended.

A roboticist may accompany the head/s of staff if their technical skills are required

.

The Head of Security has no power over the AI Upload. All Heads are equal in their authority over the Station as a whole, just with different duties and delegations. There's no reason the Head of Security should have more say on the AI than the Chief Engineer and the Research Director, especially considering the Research Director is theoretically trained in AI Theory, and in the absence of a Research Director, the Head of Security, in optimal circumstances, is supposed to bring a Roboticist, even if it's not forced or anything, because he's on equal footing.

[Surrealistik]

Surrealistik, there are Station Directives. Specifically, Directive 7. Regarding AI upload access - Station Directive 7

 

To clarify who, when and under what terms personnel may enter the AI upload.

In standard operation access to the upload requires one head of staff present with, approval of other members of the command staff. In non-standard operation access to the upload requires two present heads of staff. Approval of the other members of command is recommended.

A roboticist may accompany the head/s of staff if their technical skills are required

.

The Head of Security has no power over the AI Upload. All Heads are equal in their authority over the Station as a whole, just with different duties and delegations. There's no reason the Head of Security should have more say on the AI than the Chief Engineer and the Research Director, especially considering the Research Director is theoretically trained in AI Theory, and in the absence of a Research Director, the Head of Security, in optimal circumstances, is supposed to bring a Roboticist, even if it's not forced or anything, because he's on equal footing.

 

I selected the HoS for three reasons:

#1: Giving authentication rights to the Research Director keeps that role in a problematic area balance wise; doubly so if the Captain's authentication isn't required.

#2: Law uploads and integrity are most certainly a security matter; one of the most important in fact.

#3: The HoS starts with a loyalty implant (as I recall) and is thus harder to compromise.

I could see requiring the CE's authentication in addition to the Captain's instead; however, letting the RD authenticate alone is too problematic.

[Chada1]

Just give authentication to every head, Surrealistik, if the idea is used. Because what do you do when there's no HoS on Station but there is a single RD? You do nothing. There's a subverted AI and one of the only people who is actually EDUCATED on how AIs function is now unable to do ANYTHING about it. There's no logical, non-OOC reason to deny the RD access to the AI, if the RD is an Antagonist, they're meant to be hard to deal with, and there are already OOC rules to keep them from doing anything too bad.

To be honest I could see a Head of Security deciding to destroy the AI on every subversion if they were considered the sole authority on it, and that's bullshit. (Even if that might constitute an OOC problem as opposed to IC in most cases)

At the moment, every Head of Staff has authority over the AI, it only takes 2 Heads and Captain-level approval to enter the Upload at most. And Captain-level approval can be obtained by majority staff Approval, at the moment the Head of Security is relevant without being the de-facto authority over the AI, and the RD/CE aren't de-facto over it either, only educated on how it functions.

[Surrealistik]

Just give authentication to every head, Surrealistik, if the idea is used. Because what do you do when there's no HoS on Station but there is a single RD? You do nothing. There's a subverted AI and one of the only people who is actually EDUCATED on how AIs function is now unable to do ANYTHING about it. There's no logical, non-OOC reason to deny the RD access to the AI, if the RD is an Antagonist, they're meant to be hard to deal with, and there are already OOC rules to keep them from doing anything too bad.

To be honest I could see a Head of Security deciding to destroy the AI on every subversion if they were considered the sole authority on it, and that's bullshit. (Even if that might constitute an OOC problem as opposed to IC in most cases)

At the moment, every Head of Staff has authority over the AI, it only takes 2 Heads and Captain-level approval to enter the Upload at most. And Captain-level approval can be obtained by majority staff Approval, at the moment the Head of Security is relevant without being the de-facto authority over the AI, and the RD/CE aren't de-facto over it either, only educated on how it functions.

 

I'm fine with RD authentication in _addition_ to other head authentication, but at least 2 should be required always.

Second, even if there were IC considerations for giving RD access, the server has compromised on such things multiple times in favour of OOC considerations/balance. That said, requiring multiple head authentications, no matter the alert, seems pretty reasonable to me from an IC perspective.

[JKJudgeX]

I would like to see something done about this for sure, as AI subversion happens way, way too often. Between malf AI as a possibility and antags targeting the AI pretty much every round, I've grown really hate the AI altogether as a role, and would even go so far as to support turning it completely off if something like this *isn't* done, as I feel that half the time the role is abused (valid-hunting) or played poorly (being argumentative with command staff even when not subverted).

I'm not sure that the correct approach is necessarily involving the creation of AI uploads or authentication, as another way to accomplish similar goals would be to make hiding objectives from a state laws command impossible or actually require AI Chamber access to do. In general, the AI needs to be nerfed in either its subversion or in avoiding detection of its subversion. Remember that by requiring additional authentication, undoing a traitor or killed head's subversion of the AI becomes incredibly more difficult and unlikely.

[Scheveningen]

Okay, I think it should be fairly simple.

Authentication: requires two heads to swipe on the AI upload console, then it unlocks for the next law upload. Any combination of command staff can do this, although I expect whomever is uploading the laws to have a degree or training in regards to uploading proper programming logic to AIs.

Authentication can be bypassed by an emag (100% chance) or by a multitool (50% chance.) The AI is alerted to both, whether it succeeds or not, and finds out where the location of the upload console is instantly. This keeps it fair and removes stealth uploads from being a problem but forces the AI to be quick in informing folks what's going on.

[HunterRS]

Going to say this now, doing this will be completely valid hunting, due to one thing messed up command will instantly rush it, ruining the round for people. The AI is fine where it is, and I really do not think that this is needed

[JKJudgeX]

I'm not totally against the 2 ID cards idea, but, what do you do in a situation where the HoP is a traitor or is otherwise compromised, kills the captain, uses their IDs to subvert the AI, and only one other head is playing at the time. Only having 2 or 3 heads online is fairly common lately. He could even go as far as to space the IDs, knowing that this makes fixing the AI pretty much impossible... I dunno, something to consider.

[Scheveningen]

I'm not totally against the 2 ID cards idea, but, what do you do in a situation where the HoP is a traitor or is otherwise compromised, kills the captain, uses their IDs to subvert the AI, and only one other head is playing at the time. Only having 2 or 3 heads online is fairly common lately. He could even go as far as to space the IDs, knowing that this makes fixing the AI pretty much impossible... I dunno, something to consider.

 

Flaw of the system and totally doable, I suppose. To stop a HoP doing it with spare IDs, it can only be authorized by two existing round-start IDs with security encoding. Replacements won't work, for instance. Only gold or command silver would work, for instance.

Alternatively using head of staff fingerprint hash checks could work too. In this case for traitors, you can grab someone's corpse and click the upload computer with their gloves off and it'll register the cadaver for it.

Just spitballing ideas.

[Surrealistik]

I'm not totally against the 2 ID cards idea, but, what do you do in a situation where the HoP is a traitor or is otherwise compromised, kills the captain, uses their IDs to subvert the AI, and only one other head is playing at the time. Only having 2 or 3 heads online is fairly common lately. He could even go as far as to space the IDs, knowing that this makes fixing the AI pretty much impossible... I dunno, something to consider.

 

If you're a head and you manage to Solid Snake the captain before subverting the AI and get away with it, and then also manage to hit the upload, I think that's totally fair. It's harder to subvert illegitimately, so it'll be harder to reclaim legitimately, and heads typically have massive advantages as traitors (personally I think they should be all loyalty implanted in light of this and their IC position, but that's a discussion for another thread).

Not being able to reset the AI is more of a problem, but then, keep in mind that the one in the upload chamber is always authenticated; as long as you can thermite/RCD in and deal with the turrets, you'll be fine (unless the HoP dismantles that console of course). Also if authentication can be disabled via a multitool (but then it would make subversion EZ, even if it alerted the AI to the attempt) that would be a remedy.

[UnknownMurder]

I am totally against all the suggestions in this idea.

If the antagonist succeeded in subverting the AI to the antagonist by pouring enough time and work on it, then the work should be rewarded to the antagonist. There is no absolutely need to cover anti-security because for all we know is that the AI Upload is a secured area surrounded by turrets, and we ICly don't know if there's an antagonist or a rebel scum in our manifest or out in the space (mercenary) coming to us. If we keep applying security measures, the antagonist will have a harder time to get things done and fail to meet the OOC player's fun because of ridiculous overapplied layers of security. If the antagonist or command wants to waltz into the AI Upload, let them be, maybe they're doing it for roleplay benefits or trying to be sneaky?

[Surrealistik]

I am totally against all the suggestions in this idea.

If the antagonist succeeded in subverting the AI to the antagonist by pouring enough time and work on it, then the work should be rewarded to the antagonist. There is no absolutely need to cover anti-security because for all we know is that the AI Upload is a secured area surrounded by turrets, and we ICly don't know if there's an antagonist or a rebel scum in our manifest or out in the space (mercenary) coming to us. If we keep applying security measures, the antagonist will have a harder time to get things done and fail to meet the OOC player's fun because of ridiculous overapplied layers of security. If the antagonist or command wants to waltz into the AI Upload, let them be, maybe they're doing it for roleplay benefits or trying to be sneaky?

 

The main problem is that subverting the AI can easily be done in perfect stealth and with minimal risk. All you need is materials to build the Freeform or an antagonistic upload set (i.e. EZ if you're supplied and in Research).

Let's face it, it's pretty silly and ridiculous IC that anyone can remotely upload directly to the AI and change its laws without any kind of authentication, which is the system that presently exists (all you need is the upload console and the ruleset board).

Edited February 26, 2017 by Guest

[Guest Marlon Phoenix]

Subverting the AI is high risk and high reward. The AI is an almost omnipresent body and its upload is filled with turrets that stun or kill. You make it sound as if you can just waltz in, but you can't. You need to disable the turrets, get the law card, then upload it without the AI or someone else blowing your cover. After that, you then have a ticking time bomb before someone realizes the laws were subverted.

This suggestion would make it nigh impossible to subvert the AI reliably, thus firmly planting the AI as a tool for the station and one that antagonists would have to destroy ASAP rather than subvert ASAP, which I can garuntee would be a consequence.

[UnknownMurder]

Subverting the AI is high risk and high reward. The AI is an almost omnipresent body and its upload is filled with turrets that stun or kill. You make it sound as if you can just waltz in, but you can't. You need to disable the turrets, get the law card, then upload it without the AI or someone else blowing your cover. After that, you then have a ticking time bomb before someone realizes the laws were subverted.

This suggestion would make it nigh impossible to subvert the AI reliably, thus firmly planting the AI as a tool for the station and one that antagonists would have to destroy ASAP rather than subvert ASAP, which I can garuntee would be a consequence.

 

Then, the risk is worth the high reward? Believe me, I could simply waltz in with the Command ID.

[Guest Marlon Phoenix]

I know, because I've done it. That's one of the benefits of being a traitor command member.

Disallowing the AI to be subverted from the research outpost is one possibility, though, remote hackers are a thing even within some of our random events, since hackers target the Icarus' drones. A crew member being on the same wifi network as the AI and hacking it from the outpost is within the realm of reason. A better alternative would be an ability to backtrace all the AI law consoles to find their location.

[Surrealistik]

Subverting the AI is high risk and high reward. The AI is an almost omnipresent body and its upload is filled with turrets that stun or kill. You make it sound as if you can just waltz in, but you can't. You need to disable the turrets, get the law card, then upload it without the AI or someone else blowing your cover. After that, you then have a ticking time bomb before someone realizes the laws were subverted.

This suggestion would make it nigh impossible to subvert the AI reliably, thus firmly planting the AI as a tool for the station and one that antagonists would have to destroy ASAP rather than subvert ASAP, which I can garuntee would be a consequence.

 

It's actually zero risk, or close to zero risk and all reward if you're Command, or you're working Sci with the materials to build an upload and freeform.

Again, it's absolutely absurd you can build an upload anywhere and remotely subvert the AI without any kind of authentication.

Further, yes, while authentication means it's more difficult to subvert the AI, it also means it's harder to recover it too; with two ID authentication it actually does become high risk, high reward that you actually have to work for.

[JKJudgeX]

Maybe the reset AI module could be used without special ID, but Freeform would require multiple command-level swipes?

[Surrealistik]

Maybe the reset AI module could be used without special ID, but Freeform would require multiple command-level swipes?

 

That's not a bad idea; reset module can be used by anyone, non-standard modules require 2 different command-level swipes.