[HN Gopher] Air-Bus Hijacking: Silently Taking over Avionics Sys...
       ___________________________________________________________________
        
       Air-Bus Hijacking: Silently Taking over Avionics Systems
        
       Author : pjf
       Score  : 43 points
       Date   : 2024-07-04 16:02 UTC (6 hours ago)
        
 (HTM) web link (dl.acm.org)
 (TXT) w3m dump (dl.acm.org)
        
       | wucke13 wrote:
       | The observation that ARINC 429 can be tapped, and that an active
       | wiretap can alter data on the bus is of little surprise. The
       | technological challenge is not high, the bus is comparable to a
       | serialport at ~115200 baud. Considering that the technology
       | surfaced in the 1970s, it is of no surprise that physical access
       | restriction is the only means of security.
       | 
       | More interesting IMHO would be what can be done to accelerate the
       | adoption of new technologies (especially w/r/t cryptography) in
       | avionics. This is more than anything a cultural problem; How to
       | convince regulative bodies, how to satisfy processes, how to re-
       | balance the proven-in-use argument (where stuff gets more
       | favorable safety assessments when it has been used long enough)
       | vs crypto-agility (where the same thing from today just tomorrow
       | becomes insecure without changing itself, because of some
       | external discovery).
       | 
       | The technology is there, but the aviation community is not _yet_.
       | Another nice read in this domain is  "Economy Class Crypto:
       | Exploring Weak Cipher: Usage in Avionic Communications via
       | ACARS"[1, 2]. I only say mono-alphabetic substitution cipher.
       | 
       | An interesting connection of Blockchain-tech, safety and security
       | can be found in "Verifiable Computing in Avionics for Assuring
       | Computer-Integrity without Replication" [3]. Here the authors
       | leverages zero-knowledge proofs to prove to a downstream actuator
       | that its commands are indeed correct results yielded by the
       | application of the appropriate control law on the provided sensor
       | inputs. However, this work is probably at least a decade away
       | from being applicable in actual certified aircraft.
       | 
       | [1]
       | https://link.springer.com/chapter/10.1007/978-3-319-70972-7_...
       | [2] https://www.cs.ox.ac.uk/files/9693/fc-paper.pdf [3]
       | https://publ.sec.uni-stuttgart.de/reinhartluettighuberliedtk...
        
         | masfuerte wrote:
         | I don't want the cockpit to lose contact with the engines
         | during a flight because a certificate just expired. And even if
         | this particular example isn't realistic, adding security to a
         | system will necessarily add lots of new failure modes. It seems
         | likely that it will be really hard to do this without making
         | the system as a whole less reliable.
        
           | GTP wrote:
           | I know yoh mean this in a general sense, but I just whish to
           | point out for everyone that in this scenario you don't
           | actually need certificates: a MAC to prove authenticity is
           | enough.
        
             | baby_souffle wrote:
             | >a MAC to prove authenticity is enough.
             | 
             | This would work in most cases, yes.
             | 
             | I think the point that 'masfuerte' was making is that
             | simplicity and redundancy are _the_ bedrock of reliability.
             | 
             | Even if all the new hardware and software to implement this
             | new security model on avionics busses is 100% open and
             | audited by every laudable computer science / security
             | specialist, it's just _more_ to possibly go wrong. That
             | tends to change how people think about risk (weather
             | rationally or not!)
        
             | theamk wrote:
             | Not really, you still need some system to distribute the
             | shared key.
             | 
             | The paper's method suppresses original message then
             | generates btand-new one, so if the key is easy to find out,
             | then the spoofed message will have a correct MAC. Which
             | means you need a cryptographically secure method to
             | distribute the key, but you also have multiple consumers
             | and you want to be able to replace broken devices too. This
             | sounds like a non-trivial cryptosystem, and assymetric
             | crypto + certificates is one possible solution.
        
         | constantcrying wrote:
         | Why is the security of attacks against a data bus a relevant
         | topic? An attacker who has access to the data bus can attack
         | the plane in arbitrary ways, whatever he wants to do it is easy
         | to imagine a way for him to accomplish what he wants, even if
         | the bus were secure.
        
           | wobfan wrote:
           | Yeah, exactly. Just drop all packets.
        
           | wucke13 wrote:
           | For a local databus it has been relatively unimportant.
           | However, there are radio data links in/out of the aircraft.
           | The scope of data for these links is quite limited today, but
           | likely to grow significantly in the future. There
           | cryptographic properties become important, and a currently
           | missing. An example would be ACARS.
           | 
           | Also in some aircraft types these data buses are
           | unfortunately not so very hard to access (i.e. accessible
           | from the cabin, with undetected access being even plausible
           | in some cases). So some resilience might not hurt in these
           | cases.
           | 
           | It's important that failure of critical systems is far less
           | threatening than systems providing plausible, yet incorrect
           | data. Redundancy and monitoring catches most of the former
           | but not the latter.
        
       | rectang wrote:
       | From the abstract:
       | 
       | > _This paper investigates cyber-physical attacks on avionics
       | data buses, specifically focusing on the ARINC 429 protocol. The
       | objective is to demonstrate how message injection, modification,
       | and deletion attacks can be executed, enabling an attacker to
       | gain full control over the transmitted data._
       | 
       | I wish that vehicular systems all had air-gap level separation of
       | messages, rendering it physically impossible to disrupt messages
       | to critical systems like flight controls. I suppose that's a
       | naive perspective, but in the long run it's hard to believe that
       | we won't have to resort to provably correct systems to thwart
       | attacks.
       | 
       | > _To accomplish this, we propose a method that involves
       | modifying messages on the data bus without segmenting it._
       | 
       | Can we really live with avionics platforms as a setting for the
       | same kind of perpetual arms race against attackers that we have
       | for general operating systems?
        
         | joe_the_user wrote:
         | Another question would be: "Can we create systems intended to
         | be permanently disconnected from the Internet?". Unfortunately,
         | the answer seems to be no. You can see indications in the way
         | that small water purification systems connect to the net just
         | to save engineers from going in a weekends.
        
           | constantcrying wrote:
           | >"Can we create systems intended to be permanently
           | disconnected from the Internet?"
           | 
           | Any modern jet will function without internet.
        
             | buildsjets wrote:
             | Not permanently, which is the conjecture laid out by OP. At
             | some point, there will need to be nav data updates, updates
             | to the aircraft's required systems, updates to IFE systems,
             | etc. Modern jets do all of that wirelessly. Additionally,
             | every single modern jet uploads all recorded engine
             | parameters from the flight to the engine manufacturer after
             | it arrives at the gate. Do I know what I am talking about?
             | Ref. username.
        
               | constantcrying wrote:
               | Sure, I have worked on multiple plane systems which were
               | connected wirelessly, even to the internet. (Although
               | that is something which I am extremely glad I have left
               | behind me and would never brag about in my profile name)
               | 
               | Even though, a modern jet is still able to function
               | without internet.
        
           | SoftTalker wrote:
           | There are other means of remote access besides internet. POTS
           | (i.e. dial-up modems), cellular wireless, and other forms of
           | radio are several that come to mind.
           | 
           | Internet is almost certainly the cheapest and easiest thing,
           | which is why it's used.
        
         | constantcrying wrote:
         | If it requires physical access an attacker can also attack the
         | analog systems which are controlled by the software.
         | 
         | >I wish that vehicular systems all had air-gap level separation
         | of messages, rendering it physically impossible to disrupt
         | messages to critical systems like flight controls.
         | 
         | This is just false. There is nothing in the world which makes
         | physically separating two airplane systems impossible.
         | 
         | >Can we really live with avionics platforms as a setting for
         | the same kind of perpetual arms race against attackers that we
         | have for general operating systems?
         | 
         | The comparison is false. OSs are exposed to the entire world.
         | Airplane systems require physical access.
        
           | contingencies wrote:
           | _Airplane systems require physical access._
           | 
           | ... to potentially only _one_ of the components within the
           | system, at any point in its lifetime, across the entire
           | supply chain and all build, test /verify, operations and
           | maintenance processes.
           | 
           | (Edit in reply to child: Yes, obviously "the components
           | within the system" means those actually connected, not a
           | number 3 sprocket in seat 63E's incline mechanism. You have
           | re-iterated my point.)
        
             | constantcrying wrote:
             | False. Most components are not connected to any of the
             | relevant busses.
             | 
             | And if you had control over the _specific_ component you
             | need the plane is already compromised, whether the bus is
             | open to spoofing or not is an irrelevant question.
        
         | p_l wrote:
         | Not physically disconnecting AFDX network from IFE network was
         | one of the reasons 787 got delayed, because even with various
         | Bush cuts to enforcement Boeing was told to go pound sand and
         | redo the wiring until non-avionics bits were physically
         | separated from avionics.
         | 
         | TL;DR _it 's already a standard and has been ever since
         | possiblity of sharing the networks came to be_
        
         | Veserv wrote:
         | The problem described in the paper is not what you think it is.
         | The paper effectively says: "Assume we control the wire to the
         | flight controls, then we have complete control over whatever is
         | sent to the flight controls." Not to belittle the technical
         | work in constructing a implant that can manipulate the
         | electrical signals in the wire in realtime, but the consequence
         | of such access is as obvious as it is uninteresting.
         | 
         | Not to say that physical compromise of the wire is unbeatable;
         | encryption makes it effectively impossible to spoof or rewrite
         | messages, but the wires and communication protocol are already
         | only intended for communication between trusted components (if
         | you are communicating to untrusted components then you have to
         | use something else like a data diode). The only really
         | interesting part of the highlighted attack vector is that the
         | "trusted wires" are likely not particularly physically
         | separated from "non-trusted wires" or easy access which makes
         | physical compromise at least plausible to achieve for a
         | external malicious actor as compared to physically modifying
         | one of the actual critical flight computers.
        
           | likeabatterycar wrote:
           | > then you have to use something else like a data diode
           | 
           | From what I read ARINC 429 is a one-way bus making this
           | completely redundant and unnecessary.
        
         | likeabatterycar wrote:
         | > I wish that vehicular systems all had air-gap level
         | separation of messages
         | 
         | From what I'm reading, ARINC 429 is as air-gapped as you can
         | get. It is a one-way serial protocol (separate wires for
         | transmit and receive). Only the wires that need to be connected
         | are. Messages go from->to where they need to be.
         | 
         | Unless by air gapped you literally mean "don't connect anything
         | together" at which point you no longer have a functioning
         | vehicle.
        
       | constantcrying wrote:
       | To be honest this is both unsurprising and IMO very irrelevant.
       | 
       | Spoofing a CAN or ARINC429 bus requires physical access. At that
       | point an attacker has access to the physical systems of the
       | plane, at which point the plane is compromised anyway. What he
       | uses to take over the plane is essentially arbitrary and there is
       | absolutely nothing that would give any protection.
        
         | bilbo0s wrote:
         | I was wondering why anyone would go through all that trouble if
         | s/he already had that level of control of the plane?
         | 
         | Good to have someone more knowledgeable explain that I'm not
         | necessarily crazy.
        
           | Two4 wrote:
           | Imagine a scenario where a plane is carrying a person who is
           | enemy to a certain nation state. A nation state who is not
           | above, say, using umbrella air dart guns to poison their
           | enemies with radioactive compounds - just to make it clear
           | that we're solidly in the realm of using James Bond style
           | bullshit to secure national interests in our particular
           | scenario. If this nation state were to plant a device on the
           | avionics bus that would spoof the airspeed readout and cause
           | the pilot to nose up on takeoff before the plane has reached
           | critical velocity, it would be very difficult to find this
           | device in the aftermath of the fiery crash, and also very
           | difficult to not blame this on a faulty sensor or pilot
           | error. I say fiery, because there would likely be a full tank
           | of fuel, further reducing survivability in this scenario. A
           | scary thought.
        
             | constantcrying wrote:
             | This can be accomplished just as easily by targeting the
             | analog input for the actuators or the analog input of the
             | stick. Or attack one of the other myriads safety critical
             | systems outside of the bus.
        
               | giantg2 wrote:
               | And... attacking manually gives more plausibility to it
               | being an accident rather than having a dongle attached to
               | the plane, or code potentially surviving on the system.
        
             | wil421 wrote:
             | Why wouldn't you just mess with a mechanical component or
             | simply murder someone without doing it while they are on an
             | airplane?
             | 
             | Car bombings still happen sometimes but it's much easier to
             | just shoot someone or push them off a balcony.
        
               | a3n wrote:
               | Why wasn't Prigozhin killed on the ground?
        
               | constantcrying wrote:
               | Presumably because he was surrounded by troops reasonably
               | loyal to him, making any assassination difficult.
               | 
               | That he was killed by manipulating the data bus seems
               | entirely speculative and exceedingly unlikely. If you
               | have that kind of access to the plane you can plant a
               | bomb or if you just want to down the plane Russia has
               | potent anti-air weapons which trivially can take down a
               | sub sonic passenger jet.
        
             | Veserv wrote:
             | Maybe if it crashed in the ocean. If it crashed on land,
             | then they would almost certainly find the device in the
             | wreckage as in the Pan Am 103 bombing [1] where they found
             | even scraps of disintegrated clothing in the suitcase
             | carrying the bomb. The flight recorder would also almost
             | certainly show the nonsensical inputs and outputs and the
             | pilot confusion. And, unlike other industries, aviation
             | does real root cause analysis to identify every factor
             | involved in a crash so it would be exceedingly unlikely
             | they would throw their hands up into the air and just blame
             | the pilot or something as stupid like that.
             | 
             | The chances of a attack like this being undetectable are
             | exceedingly low. You would likely need to compromise nearly
             | every aspect of the plane to make sure you have suppressed
             | every available cross-checking mechanism. Does not stop it
             | from happening, but it would not, in any way, be some sort
             | of magic assassin weapon.
             | 
             | [1] https://en.wikipedia.org/wiki/Pan_Am_Flight_103_bombing
             | _inve...
        
         | chinathrow wrote:
         | Right, a simple wire cutter is more than enough once you have
         | physical access to the avionics bay.
        
           | lb1lf wrote:
           | A wirecutter would presumably cause an issue which would
           | manifest itself prior to takeoff.
           | 
           | Either that, or you must make your way into the bay while in
           | flight with said wirecutters.
           | 
           | Now, a dongle quietly manipulating enough variables to make
           | the plane uncontrollable in flight, on the other hand... (No
           | idea if that is even possible given access to this bus, but I
           | will read this paper with some interest tomorrow (As I fly
           | AMS-EWR... :)
        
             | AdamJacobMuller wrote:
             | Manipulating actual flight controls is probably hard.
             | 
             | Manipulating sensors is probably comparably easy, and, if
             | you can adjust the AOA sensors to make the plane think its
             | in level flight while spoofing altitude data, you can
             | probably manipulate the pilots (auto or not) into a CFIT.
             | 
             | It's happened on other flights where pilots became confused
             | by conflicting information (AF447 and many others) so
             | giving them consistently wrong data is going to cause a
             | consistently bad output.
             | 
             | Anyway though I deeply agree that once you have physical
             | access all bets are off. Sure, you can put a bug in which
             | causes a complicated aircraft failure but you could also
             | just put a time-delayed container of thermite or tannerite
             | in the right place.
             | 
             | I'd be much more worried about a software update causing a
             | major issue, either intentionally or even more likely and
             | more worrying, accidentally ALA 737Max.
        
               | constantcrying wrote:
               | >Manipulating actual flight controls is probably hard.
               | 
               | Why? The flight sticks have an ADC, just control
               | current/resistance there to get the desired digital
               | values. Of course assuming fly by wire an symmetrical
               | input on both sticks.
               | 
               | Probably far easier than manipulating the digital data
               | downstream, which is subject to lots of checks.
        
           | BiteCode_dev wrote:
           | With timer of 35 minutes.
        
         | Vox_Leone wrote:
         | >>Spoofing a CAN or ARINC429 bus requires physical access. At
         | that point an attacker has access to the physical systems of
         | the plane, at which point the plane is compromised anyway.
         | 
         | I gather a perp only has to access the right piece of equipment
         | in any phase of the logistics.
         | 
         | These buses are employed mostly in the avionics industry, but
         | they are also utilized in ground vehicles, weapons systems, and
         | other commercial and military equipment industries.[1]
         | 
         | [1] https://www.logic-fruit.com/glossary/arinc-429-guide/
        
           | numpad0 wrote:
           | That sounds like saying that one could potentially launch a
           | nuke by walking into an operational nuclear submarine and
           | pressing buttons in the right order.
        
         | a3n wrote:
         | I think it's better to take away options, rather than throw up
         | your hands.
         | 
         | Just because someone can load up my car on a flat bed tow truck
         | doesn't mean it's pointless for me to have locks on the doors
         | and ignition.
        
           | constantcrying wrote:
           | >I think it's better to take away options, rather than throw
           | up your hands.
           | 
           | No, wasting time, resources and money while increasing the
           | inherent complexity and risk of a system to gain absolutely
           | no benefit is a very bad idea.
           | 
           | >Just because someone can load up my car on a flat bed tow
           | truck doesn't mean it's pointless for me to have locks on the
           | doors and ignition.
           | 
           | Idiotic comparison, which makes me think you are just totally
           | disingenuous. The point I made was that with the same amount
           | of effort a plane is compromised, with or without a secure
           | bus. This is _fake security_ , it doesn't protect anything.
           | 
           | Just tell me an attack on the bus of an airplane which
           | couldn't have just as easily been performed outside of the
           | bus.
        
       ___________________________________________________________________
       (page generated 2024-07-04 23:01 UTC)