https://homeostasis.scs.carleton.ca/wiki/index.php?action=history&feed=atom&title=Operating_Systems_2019W_Lecture_23 2026-04-05T17:00:50Z Revision history for this page on the wiki MediaWiki 1.42.1 https://homeostasis.scs.carleton.ca/wiki/index.php?title=Operating_Systems_2019W_Lecture_23&diff=22326&oldid=prev 2019-04-09T00:19:03Z

<p>Created page with &quot;==Video== The video for the lecture given on April 8, 2019 [https://homeostasis.scs.carleton.ca/~soma/os-2019w/lectures/comp3000-2019w-lec23-20190408.m4v is now available]....&quot;</p> <p><b>New page</b></p><div>==Video==<br /> <br /> The video for the lecture given on April 8, 2019 [https://homeostasis.scs.carleton.ca/~soma/os-2019w/lectures/comp3000-2019w-lec23-20190408.m4v is now available].<br /> <br /> ==Notes==<br /> <br /> &lt;pre&gt;<br /> Lecture 23<br /> ----------<br /> <br /> Systems security<br /> <br /> - not crypto(graphy)<br /> <br /> Cryptography is amazing technology, but it is very brittle<br /> - almost nothing is proved secure<br /> - implementation is very very tricky and it is very easy to make<br /> a mistake that undermines all security guarantees<br /> <br /> Debian openssl predictable PRNG flaw<br /> - someone was trying to get rid of valgrind warnings<br /> - the uninitialized memory was used to gather entropy from the OS...<br /> <br /> Can we make perfect software?<br /> - if you can prove it correct...maybe?<br /> - proofs can be flawed and can make false assumptions<br /> <br /> When we find vulnerabilities, we&#039;re engaging in a process that never ends.<br /> <br /> But any vulnerability could undermine a whole system&#039;s security<br /> <br /> On Linux, the &quot;trusted computing base&quot; (TCB) is<br /> - Linux kernel<br /> - bootloader<br /> - every process running as root (started at boot or setuid root)<br /> - some partially privileged processes/executables<br /> <br /> Security standard practice is that we want as small of a TCB as possible<br /> - but this isn&#039;t practical, at least with current development practices<br /> <br /> How can we have security with flawed software?!<br /> <br /> I believe this is possible, because other systems are highly flawed yet<br /> are reasonably secure: living systems<br /> - we all have a variety of imperfect defenses<br /> - but they work well enough to keep most of us alive most of the time<br /> <br /> Modern computer antivirus is like requiring a vaccine against every virus<br /> <br /> Living systems use a combination of<br /> - barriers &amp; general defenses<br /> - adaptive defense<br /> - diversity<br /> <br /> <br /> The adaptive immune system<br /> - notices damage &amp; other strange behavior<br /> - tries many possible solutions<br /> - ramps up solutions that work<br /> <br /> Say I want a system to detect malicious system calls<br /> - if I do large-scale learning (monitor milions of systems, get their<br /> system calls, look for bad ones), I will face certain fundamental problems<br /> - you have to stay small if you want the accuracy to be usable<br /> <br /> My strategy<br /> - monitor locally<br /> - do stupid learning (table lookup)<br /> - don&#039;t do anything too dangerous in response to detected problems<br /> (don&#039;t kill processes or delete data)<br /> <br /> For example, slow down unusually behaving processes<br /> <br /> But the understandability problem is fundamental<br /> <br /> &lt;/pre&gt;</div>

Soma