4 ms·

Sorry, I forgot to emphasize that after that 2018 Ars article was written [1], the security of their machines has improved, I think. The wiki article has more d

• ricardobeat 3y ago
  That paper is from 40 years ago and things have evolved quite a bit since. For example, the entire toolchain that compiles the code running on the voting machines is compiled from source, in a single machine in an air-gapped secure room, and digitally signed. The attack you hypothesized, where a hidden radio “flips a bit” to alter votes, would necessitate defeating several layers of security and months-long auditing cycles. Software and hardware security aside, randomly selected devices are sent to three major public universities for verification every year. On election day, a significant number of random devices are used in tandem with paper voting, which would also flag altered results. Transparency could definitely be improved by allowing individual verification of votes, but that comes with the risk of voter identification as in the article you linked to earlier. So not really “in principle, extremely justifiable”. It’s just such a remote possibility and you’d need some kind of evidence (which he did not have) to point to a potential failure. All that said, some of his tweets literally said “wake up, the elections were a fraud” and not just innocent questioning. He used that to imply that any actions by the justice court were illegitimate. And then he went on to say the country should have a Nazi party… kinda difficult to sympathize with.
  • jevoten 3y ago
    > The attack you hypothesized, where a hidden radio “flips a bit” to alter votes [..] Software and hardware security aside, randomly selected devices are sent to three major public universities for verification every year. Sorry, I wasn't clear enough - the whole point of using a radio is to selectively compromise only machines used for voting, and leave the verification ones alone. > but that comes with the risk of voter identification as in the article you linked to earlier. The court simply asserted that a paper trail risks voter identification, but the many countries that use paper voting without problems of voter identification shows this assertion to be false. But it's easy to determine one side is in the wrong, when you simply accept the other side's assertions at face value. > The attack you hypothesized, where a hidden radio “flips a bit” to alter votes, would necessitate defeating several layers of security and months-long auditing cycles. No, you have to defeat only one layer - hardware. Put your malware into some spare non-volatile memory, and only activate it when you get the signal. Otherwise simply report that memory as empty. You can even allow writing to it (assuming you're reporting the total memory available accurately) in case the legitimate code needs more space. In that case, the attack disables itself to avoid discovery. Then all your digital signing and compiling from source (This uses a hand-written-in-assembly bootstrapping compiler, right? Otherwise the trusting-trust attack still works.) is moot. Finally, all of this depends on trusting the people doing all this verification and development. What is the minimum number of people one would need to compromise, to compromise this entire process? Let me remind you that after all the verification and bootstrapping and code signing and vulnerability patching, to an external observer, a compromised and a trustworthy machine look identical [1]. Given that the courts twice overrode paper trails, why should one trust them? > And then he went on to say the country should have a Nazi party… kinda difficult to sympathize with. Oh that's easy then - prosecute him for being difficult to sympathize with, not for questioning the election. To reiterate - a paper trail would render all these attacks ineffective. And it is proven to preserve anonymity, as evidenced by the many countries using it. Instead, a process is used that requires trust in an appointed (by who?) group of experts, whose work has to be monitored for its entire duration to be trusted, you cannot simply look at the resulting artifact, and even if you do trust the current batch of these experts, they are still vulnerable to outside attackers. The US is on record exerting behind the scenes pressure to claim the election is secure (and also provides hardware for that election), courts have blocked attempts to make it more transparent and verifiable, and questioning it is illegal. But this is all fine because hey, probably this time it wasn't subverted, and the people that don't like the process are unsympathetic, and conspiracies never happen. [1] No, dumping memory won't help - unless you decap the memory and read the individual bits with an external tool, the machine can simply lie about the contents of its memory.