2018-07-17 4:07 GMT-03:00 Norbert Hartl <norbert@hartl.name>:
Am 17.07.2018 um 05:14 schrieb Hern��n Morales Durand <hernan.morales@gmail.com>:
2018-06-27 23:21 GMT-03:00 Otto Behrens <otto@finworks.biz>:
Thanks for the effort guys.
I tried to download the image, sources and vm separately (basically
extracted what https://get.pharo.org/64/61+vm does), but ran into fresh
trouble.
Firstly, is "wget -O - https://get.pharo.org/64/61+vm | bash" not risky in
terms of security? It should be quite possible to inject a lovely trojan
horse with this, or not?
Yes, this is possible:
https://www.idontplaydarts.com/2016/04/detecting-curl-pipe-bash-server-side/
I like this kind of articles because it proofs every time how less people know about security. If it comes to security eyerone goes hysterical.
The basic assumption in it is that you need back pressure in order to detect it. But as the get.pharo.org script is not big enough.... the whole thing is not possible. Finished. Next.
Well, you have a case: "People telling people to execute arbitrarycode over the network" : https://curlpipesh.tumblr.com/ ;)When it comes to security, it is better to keep humble. Even the guysbehind cryptographic functions got caught, so it is better to followthe best practices around, at least in a really complex domain assecurity.You don't need a big script to get fooled, have a look athttp://people.zoy.org/~sam/filsdepute.txt and copy-paste its contentsin a text editor.But what is really funny is that there is something not mentioned. Because for this to work you need to have control over the server meaning get.pharo.org. If this is the case I can also make the client download a different image with malicious code or a vm or ..
Or even better. Every piece of software you download has the same problem. Just because it comes with an installer doesn���t mean it is safe.
Security problem scenarios are often theoretical problems. The need to be checked against real conditions in order to identify a threat or not.
We don't agree this time Norbert. Even without a real scenario thetheoretical problem is enough. Two examples are: 1) In security if avector is prone to attacks, with a success probability, then could beconsidered obsolete or weak, and weakness increases over time (SHA-1,SHA-2). As computer speeds gets faster, they just calculate analgorithm resistance and estimate its life-expectancy. 2) People usepackage managers precisely because is a delivery mechanism whichtransfer the burden of trust from the Pharo maintainers to the packagemanager (which verify signatures). That without counting the pleasureof deniability if an installation got compromised :)
I���m not sure why we disagree here. I said that all of those scenarios need to be checked against real conditions. An algorithm that can brute forced due to the increase in computing power is a real condition. What I was saying is that not every theoretical problem is one you need to take care about. Why I am saying this? In my career I met far too many ���experts��� dealing every theoretical problem as a real one without understanding the impact. Those people often sit at important positions in corporate infrastructure restricting the usage of the internet and software to a bare minimum. And they cannot see that they produce an economical damage to the company while keeping everyone from working. In order to justify that the damage it should at least be a real threat, don���t you think?
The cost of the measure you take to protect needs to correlate in a sensible manner to the value you want to protect.