This is the Agent Smith Army problem:

A hacker can write his own agents and programs with a few months downtime. He now has the source code and can create unlimited copies for free with no effort. At chargen, a hacker can buy 25 top of the line commlinks with top of the line OSs. If the criminal doesn't want to pay for the commlink, he can rob a whole store full of them. If he doesn't want to rob a store, he can do what hackers do today and hijack thousands of nodes throughout the matrix and install an agent on each of those. (It's called turning computers into zombies).

A hacker can only control system * 2 agents at the same time, but he can pre-program them with their instructions before stashing them. The hacker now has thousands of agents running all over the world ready to attack anyone he's programmed them to. Since they are each running on their own node with a decent system rating, none of them suffer degradation even if all one million of them attack one node.

Since defenders can do a similar strategy with their IC, Shadowrun goes from street level to Lord of the Rings level with armies of thousands of agents clashing in commlinks.

Solution: Agents always degrade whatever node they are on, they do not have a source node, and their icons do not project outside their source node.
Explanation: If you say that agents can not project their icons across the matrix as a hacker can project his persona and say that an agent always runs on the node their icon is in, then agents and IC cause degradation on whatever node they are in or are attempted to gain access to. An army is therefore self limiting because after a few agents with a few programs, they have their Response reduced so bad they can't even move.

Clarification: a hacker's persona still originates on his commlink and his persona never suffers degradation when on a low-system node or the presence of an Agent Smith Army on a node. His persona is only degraded by programs running on his originating source commlink. The hacker can dance through a node that has several high-program ICs bogged down by their own weight.

Question: Should a technomancer's sprite's complex forms also suffer this degradation? Do they get an unfair advantage over hackers if they don't? Is it ok that TMs have an unfair advantage over hackers in node hacking given a TM's other shortcomings?

Does this solution fix the Agent Smith Army problem? Does it cause other problems?

How many agents does it take to do an effectively complete denial of service attack? Particularity if you are abstracting to to be the NETWORK and not the NODE.

I suspect much fewer than what a small mediocre street gang could put together when they decide to stop dealing BLTs and decide to go into the matrix blackmail business.

I think you are over complexifying the both the problem and the solution.

No other program has any association with the number of entities that are running. 'Browse' 'reality filter' 'black hammer' - it doesn't matter.

Even pilot don't matter - pilot is just the quality of the abstracted system managing a drone

The only programs in the game that do not operate in this totally abstracted service paradigram is *agents* and IC.

The fix is just to move agents and IC into the same paradigram. So multiple agents with the same/similar orders are just treated as one agent, and multiple IC on the same node are just treated as one IC program when defending.

So if you get 10312531513414 agents and task them to smack down renaku's regional office - you just made a rating 7 agent! Grats!

Solution #2: You could say that (for some reason) agents need a hacker's persona to function. Agents can only run from a hacker's source commlink, just like the hacker, and therefore are limited by the commlink's response degradation just as the hacker is. You can have as many agents as you want and the agents can go anywhere they want but they can only run from the hacker's commlink. This limits an Agent Smith Army because a commlink can only run a certain amount before all their programs start to degrade the hacker and all the other agent's responses. You could apply the same principle to IC by saying the IC have to spawn from a security hacker's server. You might even say that an agent/IC's rating itself counts as a program for the purposes of system degradation.

I have no idea why an agent/IC would need a hacker's persona, maybe it needs to use part of the human brain for background processing or something?

We definitely need to say that you can only be jacked into one commlink at a time. A metahuman connects to one and only one commlink via datajack/skinlink/wirelesslink and views the matrix as AR or projects his persona out into the matrix via VR.

So far, I like solution #2 the best. It's the most elegant but the least plausible.

For plausibility, I'm thinking that we say that a Pilot can run on drones, vehicles, and gun platforms without human assistance without any problem, but hacking in 2070 is far more complex than flying a helicopter or firing a rocket launcher. We could say that the extreme complexity of future hacking requires the Agent/IC to tap into a metahuman's brain through the commlink interface so the Agent/IC can utilize the background processes of the metahuman brain for its advanced decision making and cognition tasks. That's plausible enough I think and I think it completely solves the Agent Smith Army problem for both the attacker and the defender.

Anyone spot flaws?

Falling asleep just now just gave me a great idea. We could say that an agent can tap into a metahuman's brain's neural net even when while the metahuman is asleep and can therefore keep running 24/7. That explains the "unmanned" systems that are only populated with IC and no security hackers. The security hackers are somewhere, just off duty, probably asleep, and aren't paid to pay attention or get involved, just act as a host for the system's IC. Anyway just a thought, back to bed.

Or the player and GM come to a tacit understanding at the begining of the campaign that the player won't create armies of agents and the GM won't create armies of IC...

Simmilar solution for the Bloodzilla problem. After all remember, anything your PC can do the GM can do better.

The implicit assumption here is that armies of agents is somehow not "street level". In fact, it is. It is no less street level than a street punk who occasionally plays a Throalian courtier in Age of Atlantis is.
Agents exist solely in a computer. They aren't physically real. Thousands of something that isn't physically real is no less street level than one of something that isn't physically real is.

The problem with agent armies is that the size of a runner's agent army is totally arbitrary, thus matrix battles are decided solely on the GM's arbitrary choice for the size of his agent army.

The simple solution is actually the best solution. Agents cannot be copied. It is simple. More than that, it provides an IC explanation for why gigantic agent armies of arbitrary size don't exist. Every Agent must be coded separately and every Agent must be purchased separately.
Under this house rule, any agent army is limited in size by the number of man-hours its owner can dedicate to coding or the amount of money it can spend buying the fruit of other entities' man-hours and the total number of agents in the world is limited by the total number of man-hours that all coders in the world combined can work.

This is supplemented with the rule that Agents cannot code other agents.

alternately, you could rule that agents actually require physical hardware (which, coincidentally, has a parts cost equal to half the cost of an agent, with the fully assembled thing costing as much as an agent) and that therefore you have to actually buy each agent. this also has the (very important imo) added benefit that a hacker cannot steal an agent through the matrix. of course, they can still (presumably) travel around like a persona, but it will be a lot more obvious to the owner that something is going on... so you could hack an agent/IC to do something for you once, but after that it will be pulled offline, and modified from beyond your reach (unless you physically enter the facility and steal the hardware, of course). they would still require something separate to operate on, of course.

truth be told, this rule (or something like it) should probably be used for pilot programs as well. at least this way the hacker has to actually spend *some* money to be able to copy/steal agents.

Actually the Matrix is abstract enough right now that you have tons of solutions. Some may be better than others, but in the end it depends on how you consider the Matrix.
My solution might use concepts that you don't want to see in your Matrix.

For example some people are perfectly fine with high rating agents being as capable as hackers ("hacker-in-a-box"), while others will think that agents shouldn't be able to replace characters.

Oops, double post.

Your proposed solution of having agents deteriorate the target node makes offense win. Drop a few agents on any system to lower response to one and then waltz in personally and own the system, due to still running all your programs at full rating.

Cthulhudreams is on the right track.

Alternative idea for multiple agents trying to hack their way in to a node -

Since they're all using the same code, and all going up against the same defences software is likely to make the same decisions on both sides thus you just roll as if it was one vs one.

Or whoever it is that runs the matrix links chases you down for spamming out the system.

I know there would probably be ways to randomise certain parts of it, but why let things get more complex than they need to be for the average game? Yes I hope there will be a rules clarification or errata but you don't need to work out *every* problem in rules terms, just come to an understanding that if someone tries the army method "bad things ™" will happen. Saves everyone headaches. Unless of course you're the rules lawyer type who enjoys this sort of thing.

I suggested this in another thread, but it seems to have largely gone unnoticed. To sum up (or try): Identical agents cannot be subscribed to nor present in a node at the same time. If you are concerned about power creep as your hacker codes one new agent after another, implement SOTA rules.

Would this work?

Likely not - people are searching for a hardcoded solution rather than a soft "to expensive". And there should be no need for differentiation between programs of the same type and rating.

One of the easier solutions is limiting agents to rating 3, maybe 4. Those dicepools amount to nothing against secure systems, and reduced load hampers most tasks that are considered a problem right now. Most "legitimate" tasks are still possible if the agent is custom-loaded for the problem at hand.

Or, you could just have agents work exactly the way they do now, and the GM says, "No, your not doing that." That solves the problem and you do not have to change any rules to do it.

I never liked the idea of the GM just saying "if you do that, Thor the god of Evil GMs will strike you down." It's better to have coherent and consistent rules, not GM fiat. I don't think this is a hard problem to solve, there are several good solutions out there.

I don't know if you can say that identical agents behave identically, because we see that they don't. I'm not sure how to justify how a million of them is no more effective than 1 of them. What is the cut off and how do you justify the cutoff?

Reducing their effectiveness to rating 3 or transforming them into teamwork enhancers or smartlinks is also a good solution.

I don't think you can say that the hacker can't just make copies of the agent because the agent is pure software. Software can always be copied.

I had once suggested in another thread that you could say that an agent requires some sort of hardware component as Jaid suggested here. For example, the famous arbitrarily-difficult-to-reproduce "neural matrices" of Star Trek: the Next Generation's androids and Voyager's holopersona. Star Fleet Directive #9, justify the plot. If you said the agent's hardware component costs a couple thousand nuyen that would put a stop to thousands of agents, but even a starter character could still have at least a hundred of them. You'd have to bring the price up to at least 10000 nuyen so runners could never have more than a dozen of them. Companies could still have armies of agents but it would be hard to justify because if they had that kind of cash laying around for security they could also hire armies of hackers.

IC Operates on the Node it's in as per RAW so no army there, Sprites are pretty limited unless you bring an army of Techno's. So BIG problem is simply Agents. Why not a simple solution that clicks easily into existing rules, no waveing it off or creating whole new set of rules.

Simply limit number of Agents that can run independent on a Node equal to Nodes system rating. (This can simply be a matter of way Nodes hardware is set up and wouldnt effect *search functions.)

*Would still allow agents in AR to use basic common search functions on public Nodes without the penalty as described RAW

Example: Node with System of 5 would let five Agents running on there own Comms or the hackers comm be active in the Node. This would'nt effect agents running on Node in question which would degrade Response as normal.


Sound reasonable?

I go with the explanation: You can't have multiple copies of an agent running in one node. The Matrix was built from the ground up with the prevention of self-replicating code in mind. It's impossible, the GM has waved his hands. If you want two agents in one place, they have to be two different agents; two completely different programming tasks to create. It doesn't have or need a justification in modern computers, because it uses SR computers that run on 80's concepts of awesome.

If you want to have multiple copies of an agent in multiple places, I really don't see it as much of a problem.

Otherwise, I allow agents to function just like hackers: They have a node on which they run and determines their response, and up to Ratingx2 nodes on which they have an active persona, just like a hacker. Sometimes, it'll be the same node.

So far I haven't had any problems, YMMV.

Like idea of not allowing more then one "copy" of an agent in same node at same time.

I don't think having security hackers chase a PC hacker if they're utterly spamming out the system, or at least freezing their link as close to the course as possible, is striking down a player with GM fiat. Using an army of agents is about as subtle in the matrix as a squadron of Rotodrones with AK97s flying around a busy shopping centre is in the meat world, why wouldn't somebody intervene? If not the company that runs that link getting mad and trying to keep their system from being slowed and bringing a portion of the local matrix to a near stand still then maybe law enforcement intentionally trying to stop this type of attack.

I also like the anti-replicating code idea. Either, or even both, are plausible.

This solution doesn't do much for the IC army aspect but as for Agent Smith....

If we assume AS attacks are over kill for hacking a PAN, then any network that an AS army tries to hack has to go through the first node of a network connected to the matrix. When the AS army enters this node the added traffic has to be noticed alerting someone on the security side. Why not then have the node shut down? This doesn't effect nodes operating on the network behind the matrix node or compromise their security at all. It also limits any damage a AS army could do because they lack the ability to change physical location and try to jump into a different node. At this point security knows they have a missive assault attempt on the public node and begin tracking down the sources of the DDoS attack, landing the hacker in some serious trouble. Yes the corp has to route traffic to another network to support remote users but as the AS army follows this increases their chances of being traced and taken care of.
Agent Smith armies are not a subtle way to hack and really run counter to what I image the goals of a Shadowrunning hacker to be. If the goal is to shut down matrix service and a hacker is willing to invest the time and money into the creation of an Agent Smith army then I say bring it on. They should be prepared for the IC army counter and a gloriously fought Lord of the Rings style showdown! One comlink to rule them all

did anyone poke a look at mye quick corp host writeup that i did?

edit:

heh, it had already dropped of the first page. talk about lack of interest.
here is the url, just for kicks:
http://forums.dumpshock.com/index.php?showtopic=18863

thr true irony of this is that the agent smith 'problem' is almost identical to one of the most common real life hacking methods the DDOS

the major difference is that the bots doing the DDOS are much less intelligent than the ones in SR

There's the Matrix in the sense of the large number of interconnected nodes, in more or less the same way that the World Wide Web involves lots of interconnected computers (and arpanet did long ago on a different scale). However, whatever matrix rules you use, ought to also be viable if two people on an isolated facility in Antarctica hook up their two comlinks and one of them tries to access the other's private diary. Or if one of them has a stack of comlinks and tries the same. Or if one of them custom-builds a comlink that doesn't have the "standard" built-in limitations.

It seems strained at best to rule that it's impossible to type orders into a stack of comlinks, each running an agent or two. Ruling that most systems are designed to detect self-replicating AI, and designed to send an alert then shut down when they detect AS, is reasonably plausible. So a hacker with AS can trigger shutdowns, but can't usefully change or copy a well-secured system's data with AS.

I've just been making a higher powered character that is my want for a game on the boards here, and i've noticed that the agent smith problem is in fact far more low key an insidious in addition to the huge over the top stuff here.

For example, just one agent is as good a hacker than my logic 9 character, and each of the guys drones throw 8 dice for electronic warfare, pretty much by accident.

There's always the simplest solution: no agents.

It makes sense, from my perspective. IC costs the same as an agent, but must be immensly simpler to code. Their orders are:
1) Find something that doesn't belong.
2) Kill it.

Agents have a much tougher job...they have to perform whatever random task they are assigned. It's easy to imagine that such a program would simply be impossible to code.

The Hackastack is a problem related to but distinct from Agent Smith.

Simply, you can hack in complete AR, meaning that there's no actual connection between you and the computer you are using to hack with. So you can simply have multiple personas in the Matrix. Each commlink you have in the hackastack has the same relationship with you as the one you are "supposed" to use, so it should work normally.

What this means is that Black IC is completely meaningless to you because you've chosen to not allow a direct connection between your brain and the commlink (so you can't take any damage). And also that Matrix damage is pretty unimportant as well since getting your persona crashed would just make you switch to the next commlink in the hackastack (which isn't even an acion, because the other commlinks are already operational).

-Frank

I don't recall any text that says that a matrix program has a single "whole" icon.

I could easily abstract a rating 5 IC as a pair of wolves, or a horde of rats.

Likewise, having a half dozen agents storm a node should be abstracted as the rating of the original agent. My justification for this is that the agent's rating takes into account its ability to cooperate with itself and other programs. Loading multiple copies of an agent is just throwing more of your hardware resources into the same algoritm -- the game abstracts anything that can use higher rating hardware as a higher rating program. Plus, having 10 identical programs works nowadays because of denial of service -- but in an age of "unlimited" bandwidth, there is no such thing as a DoS attack. Instead, you have 10 programs all attempting to access the exact same weakness on your own icon, and all with the same identical weaknesses on their own icons. It shouldn't add anything to use more than one.

My problem with Agents is that the default starting agent (rating 4) is nearly as capable as the typical default starting hacker (Electronics Group 4 / Hacking Group 4), and costs far less in BP. Edge is the only difference. (Incidentally, agents also make hacking skillsofts useless! You can just use an agent instead, and it'll cost less too!)

10 agents attacking a node is identical to having 10 individual human hackers attacking the node. Agents work like people, and they are expert systems able to do a very good job at whatever they are designed to do. If you want to assume that only one hacker can attack a node at once, I'll suggest that KE will offer a hacking service in which one totally useless incompetent hacker will attack all their subscribers 24x7x365 to keep out the competent hackers.

That is the very problem - the agent is not really limited compared to its controlling hacker. Until the hacker gets augmented, it will easily have better dicepools. And it offers deniability.

The brain computing idea of Frank is a neat approach, requiring a connection to the users brain. This can also be used to limit the number of agents (mechanic for bound spirits essentially).

Then you need to remove the advantages a rating 6 agent has over moderate hackers. That can be done by making any agent use the hackers skills, as we just mandated a connection to the users brain anyway. Now it is an agent of your brain, and technomancer sprites have some advantage in that they can go independant if need be. If one cares about TMs.

Then you need to say that teamwork tests are beyond the coordinative capabilities of the agents and you are done. A few, usually one, agents per user, abilities scale with the users power level (mages get that all the time), high-end advantage for the hacker due to augmentations.

If anyone's interested, I've got some house rules trying to stay close to the canon while expanding both fluff and rules. There are some fixes for Agent Smith (actually there are rules for doing an Agent Smith attack) and Hackastack and other troubles.

I've been working on the Matrix nearly since SR4 came out, so these aren't just quick and dirty fixes (I hope).

Here they are, just be sure to check page 2 as it has some more details and rules:
http://forums.dumpshock.com/index.php?show...opic=19782&st=0

Blade, your system doesn't fix Agent Smith at all.

Remember what Cthulhudreams said about the subtlety of the Agent Smith problem? Where he could run electronic warfare on his drones and outmatch all but the most powerful hackers with each and every walking toaster in his van? Where his death agents were rolling more dice than most characters could have? Right, the only thing your rules do to agent attacks is to give defending systems the ability to arbitrarily jack their hacking defense up to infinity if "too many" hack attacks go on simultaneously, which neatly stops the one hundred billion Agents problem.

But it doesn't do shit to the four Agents problem, which is just as game destroying.

Also, giving people the ability to jack their hack defenses up to infinity under any circumstances is really obviously bad precedent in a game where hack attacks used by player characters are supposed to accomplish something.

-Frank

Thanks for your opinion Frank. But next time please make sure you read and understand everything before posting it because your reply has little to no relationship with my system.

I'm impressed by your inability to divert a bit from your mindset.

Agents you buy at the start of the game ARE NOT independant agents. So you can spend as much BP you want you won't get any independant agent because they're not for sale anywhere.

On page 2, there's something about independant agent swarm that explains how it works to get an independant agent: it's exactly as getting hold of a Nuke or a Thor system.
There are no rules in SR4 for getting a nuke, they aren't in any equipement list and there are flavor explanations (well actually there aren't but I guess that a little bit of common sense should be enough) to why it's the case.

The nature of the hackastack problem is such that you can't solve things with restrictive agent protocols. It's literally impossible.

A "Persona" in this case can be generated by me holding a commlink and pressing buttons on it. And therefore it can be generated by a crawler drone holding the same commlink and pressing the same buttons. And therefore it can be generated by a computer program which tells the commlink that the same buttons have been pushed.

As long as "I" am distinct from "My Persona" on the Matrix through an actual air gap. So long as that is even theoretically possible, the Aget Smith paradox exists.

And all the stuff about the limits of how Agents can interact with things doesn't mean anything at all because they can launder their existence through an air gap and then it's back to buttons being pushed - same as any human hacker pressing the same buttons.

---

It is not conceptually workable. All your tirades on the subject are just obfuscation. All your base are belong to us. Us in this case is 12 - or even two - aztechnology crawlers with a Pilot Rating of 6. Caps on how many login attempts can be made don't fix the problem. Limitations on direct agency effects don't fix the problem. So long as equipment on a shelf can make a series of attempts to crack a system which are in aggregate more likely to succeed than a single character with massive BP expenditure in Matrix Know How, the matrix specialist archetype is dead.

And despite weasel words like "The problem of that method is that due to the difficulty for an independant agent to exploit a node (see opening post), there's a good probability of failure." - the rules don't back up your assertions. You may have convinced your players to not try to break the game - but you haven't made the game not be broken.

-Frank

Actually this is also covered, but if you won't read (or understand, or accept , or whatever your problem is) it, I won't lose my time explaining things to you.

Forcing people to purchase fake sins for extra commlinks does not do what you think it does. And since all your rebuttals are just "Nuh-uh!" and "Why don't you read!" then I can't say that I am terribly sorry that you are giving up on this conversation. We aren't really having a conversation. I'm pointing out structural flaws and you're pretending they don't exist and not explaining your position in the slightest.

-Frank

Ok, if insist:

1. Independant agents are just like Matrix nukes. It's hard to get one, it's hard to build one yourself and it's highly dangerous to do either. That's why there are as many rules covering independant agents as there are rules covering nuclear weapons.

2. Using a drone to control a persona is asking an agent to control a persona. So it's exactly as having an independant agent. So it'll be blocked by the security protocols of the Matrix. I don't explain what they are, because if I knew how to check if a user is an advanced agent or a human, I wouldn't be writing this but selling the ultimate Captcha solution. Anyway, what's important is that there are security features in the Matrix that prevent agents from accessing the Matrix as personas, and it will also apply to drones accessing the matrix because it's exactly the same thing.

3.