Agent Smith Problem, The problem and how to solve it Options

[Serbitar]

A Question Frank: Arent you just pushing the problem to another level?

After all, an agent is just a hacker in a box, and why stops an agent (conceptually) from getting influence as well as a hacker does?
Whats so special in a hacker that only a hackers commlink can orchestrate lots of resources, and an agents commlink can not?

If you dont want to push the escalation level arround, you will have to find a reason why an agent is not working exactly like a hacker (with maybe lesser stats).
And in most definitions I know an agent is exactly a hacker (at least in the very confined, relativley simple and artifical spaces of the matrix).

[FrankTrollman]

[Garrowolf]

okay so it seems that people don't like the idea of confining the agent.

Actually I would like to hear why people want to keep them in the first place.

I think that Frank has a good point and we need to keep things like DoS attacks as effect based. But if we start basing our hacking on influence and things that are outside of the effect of the hacker's equipment then we end up with removing them as characters. There is little to no reason to even bother shadowrunning. They could do pure hacking and not get involved in the physical side of things. They can get what ever they want from matrix manipulation. I'm not saying that I disagree with this possibility because then they become pure NPCs.

I think a distributed model for hacking would be a very interesting system. It should be based more on skill level then money at that point I think.

But let me go back to my question: Why keep agents in the first place? As a GM what good do they do? They seem to be more trouble then they are worth.

[Garrowolf]

[ShadowDragon8685]

May I propose a much simpler soloution to the Agent Smith Swarm problem?

As all Agent Smiths are just copies of Agent Smith, what works on one Smith works on all of them. So once an IC or a Security Decker has compromised one Agent Smith, he can simply take a (Whatever is the lowest division of action in a combat pass) and eradicate a Smith.

So, sure, he'll get pummeled while he's still trying to take down a Smith. But once he does, he can simply handwave away large numbers of Smiths. And worse, the whole system will know how to take down the Smiths, so IC can assist him.

So the Smiths start out as "OH MY GOD AGENTS!", but they quickly become ablative meat that can attack.


Of course, by using multiple different Agent programs, even if of the same rating, you can bypass this - Now it's not:

Hacker "Killjoy" + 11x Agent Smith

It becomes:

Hacker "Killjoy" + 3x Agent Smith +3x Agent Jones +3x Agent Johnson +2x Agent Kilroy

[mfb]

that's indescribably unrealistic, and doesn't really solve the problem. if you've got a hundred Smiths, knocking out two of them per pass isn't going to have any appreciable effect. besides, as you note, you can just have a bunch of different agents, and you're back to square one.

and how in the world is that simpler than just adding bonus dice?

[Garrowolf]

I still don't understand why we NEED agents anyway but how about this?

The book says that the firewall can increase by +4 when an alert goes off and it can shut down outside connections. So basically the first combat sets off an alert and makes it harder for the others. THen the second and third tell the system that too much is going on and shuts down the connection. Basically using more then 1 or 2 agents makes it always fail as a hack.

[kzt]

[Garrowolf]

Then why not create a Denial of Service attack program or action and move on? Actually using that many agents requires that many rolls. Have a program that just pings the firewall until it freaks out.

It isn't a question of is a DoS attack useful really but why allow that many agents and all the problems that come with it.

[Blade]

My take on the issue (I have an "open" approach of the Matrix so the mechanics described here may not be linked to any written rules):



1) Attacking
Short version : A secure configuration of a node prevents that kind of problems (just like a secure configuration of a router prevents Dos Attacks nowadays).
Long version : You have two ways to get a lot of agents :
a) Run them on commlinks, and have them linked to your persona : you're limited by the response of your commlink.
b) Run them on other nodes : they need to hack their way to the target node and then they load themselves on the node. A node set properly will have restriction to prevent overload. It can also simply refuse loading agents coming from unauthorised nodes. Of course, it can be possible for the agents to spoof their incoming adress, but if the node sends a confirmation request to the supposed source node, it will be harder... and so on.

If the hacker has got himself an admin access, nothing prevents him from modifying the server rules to load as much agents as he wants... But if the hacker has hacked an admin access, he doesn't really need a horde of agents to attack it...

2) Defending
Several explanations can be used :
2.1) Response degradation : by RAW even big nodes are subjects to response degradation, so you won't run too many agents on it.
2.2) Incompatibility/Interferences : Today, running multiple antivirus software lead to a lot of complications due to interferences between them. It might go the same way for agents. If you have 1 agent, he will easily spot the lone attacker. If you have 100 of them, they will check each other, and it will be easier for the hacker to go unnoticed (and it might generate a lot of load on the node, fill the logs with useless "Agent A scanned Agent B" lines, making it harder to notice interesting events.
2.3) Security reasons : A few agents are enough to do the job properly. Allowing for more is allowing for an "agent horde" attack, so rules restrain the maximum number of simultaneously running agents.

That's all I can think of for now.

[ShadowDragon8685]

My soloution is simple, because it means that Mr. Security Hacker can set up an Anti-Agent Agent, the system's IC can act anti-Agent as well.

So let's say he takes down one of the Agent Smiths. The Black IC in the node gets a copy of the Smith-b-Gone, and then Security Hacker sets up his own Agent, Leroy.

Leroy and the IC handwave away two agents per pass. They'll be through the MAXIMUM agents that the intruding hacker can bring in three rounds, two if the Sec Hacker helps them.

[FrankTrollman]

[mfb]

ShadowDragon8685, you have ideas about simplicity that are different from the general consensus held by the rest of the world. cherish your uniqueness.

[kzt]

[Rotbart van Dain...]

Like I said - look up Bluefrog. ;)

Things might become... interesting... when Okopipi hits the road.

[Blade]

I know that all DoS attacks can't be prevented by router configuration, but the most standard DoS attack can be. And that's what I wanted to illustrate : if there's an easy and dangerous security hole, it will be known and corrected.

If someone discovers a way to have a 50/50 Hardened armor without any penalities, you'll soon have -50AP guns/ammo to deal with that.

[Garrowolf]

which is why I think that we need to keep the system simple. There is always going to be an arms race going on. We can get caught up in a specific system and come up with alot of pros and cons, bonuses and penalties but I think that it moves us in the wrong direction.

Think about the overview of the system and paint it in the broad strokes. Then decide what we need to accomplish that with a minimum of system and rolling.

So far I have yet to hear a reason to keep them in the system at all. They seem to be a can of worms on the hacker side and a way to make infinite perception checks on the server side. One way of resolving seems to provide for inifinite rolling. Another says that the use of them would shut down things. Some people want to use them to bypass the hacker character altogether.

So out of game what is the benefit to having these things? Are people afraid to remove them and are just going to patch them up or are you just going to handle it out of game and say that you can't do certain things and ignore the in game issue?

[kzt]

[FrankTrollman]

[Garrowolf]

So the same solution works for both. Make it where agents are a function of your OS and not independant. THey can make data searches as commlink agents or they can move the drone around as an object but there is no reason for either of them to be able to move off of those systems or hack. If we make the agent and the pilot just ways to assign actions to your systems that they are normally good at but restrict them otherwise then the whole problem goes away.

An agent becomes more like Andromeda then Agent Smith. It is jsut another way of talking to the ship, commlink, or Drone. You can't send out many copies to do your bidding. They don't need to leave your commlink to do a data search. They browse the matrix just like we browse the internet today. They become a secretary instead.

Actually maybe that would solve the problem as well. If we called them secretaries then people would not try and assume the agent smith logic for them and just have them answer calls and stuff.

[kzt]

[Blade]

I get your point Garrowolf, but what of the IC ? Do you need a hacker in every node you want to secure a little more ?

[Garrowolf]

Treat IC as a local attack. If you leave the site and they feel the need they can trace you, or they can trace you while they engadge you in combat or whatever. You are already linked to their server. They don't need to send anything to you. Think of IC as a local agent on the server that uses the attack program on you. Don't think of it as something that goes to your commlink to attack you. If you are using much in the way of simsense then that simsense is sending a signal back to your meat body to feel this or that. The server is messing with that simsense signal to effect you.

There is no reason for IC, Agents, or Pilots to be able to leave their systems.

[Garrowolf]

kzt the Pilot program doesn't leave it's system to drive. It is in the cars computer which think of as a specialized commlink. I'm not saying that Drones don't move. I'm saying that the pilot program doesn't leave the drone. The IC desn't leave the server. The Agent doesn't leave the commlink.

There is no reason to have multiple Agents because they can activate as many actions in the system as you. They just are a nice interface and can do somethings that you tell them to automatically. A drone defenately doesn't need several pilots fighting for control. IC can attack as many people as it wants to.

What I'm saying is if these things are just intelligence routines attached to a system then it becomes an attribute for the system to use instead of as a bunch of robots. Treat it like a smart side of the OS. It can only do what the system can do in the first place. It doesn't need to have seperate programs or stats beyond it's rating. It becomes another rating for the computer after system instead of another set of stats to keep up with.

[mfb]

this strikes me as being non-simple. you basically have to rewrite the entire drone/agent system, rather than adding a limitation and a new source of dice. moreover, this solution makes agents basically useless, except as node defense and making searches easier. woo. we're looking for ways to reduce the effect of multiple agents, not castrate them completely.