Datasets:
Devilishcode
/
hacking

Dataset card Data Studio Files
xet
 Community
text
stringlengths
0
1.99k
(really?), let's just try to balance things a bit before it happens.
|=---=[ How did Phrack influence you and helped shape who you are?
A lot. Along with other zines, Phrack always stood out for its technical
content. I remember studying all articles on heap exploitation (w00w00’s,
MaXX’, the anonymous one) nergal’s article on ret2libc and klog’s on frame
pointer overwrite, grugq’s ELF article, and his and scut’s on ELF
encryption, and many more that I now recognize browsing the online issues.
I used to print those articles and read them over and over. I even carried a
few of the original printouts with me through many moves over the years. A
few months ago, I found the stack and finally gave them a new life.
Reading all the tricks, understanding all the different points of view,
finally helped me develop the instinct that a bit is just a bit, and all the
meaning is in the observer.
And I figured I’m not a lonely weirdo who ENJOYS squeezing the constrained
options a vulnerability offers to conquer the execution flow. We are legion.
|=---=[ What is your favourite bug/exploit?
I sadly forgot many, I feel the empty space in my memory. Let me try a few.
== CVE-2004-0368 - dtlogin double free.
Not the vulnerability, but the exploit. I was writing exploits for Core
IMPACT, and need to get it to work always. It was tricky, because getting
the double free to do a write-anything-anywhere depended on the heap state
which has to be assumed dirty, though who used dtlogin? Target: Solaris
running on SPARC. It was also a great time, because one of my friends-idols
Halvar was in town visiting, and sitting in a crappy chair at Core, working
on his bindiff, showing me early versions, and introducing me to yED (thanks
for that!). So, I had the problem of getting a reliably heap exploit, and I
started logging all traces (long life dtrace & truss), but it was impossible
to follow in text, so I hacked a GUI to show heap movements (eventually
released as HeapDraw / HeapTracer just when Alex Sotirov released his Heap
Feng Shui with an obviously much trend name). Also found how to turn the
double-free in an information leak, that allowed me to get pointers
(read-anything primitive), to finally get a very reliable exploit. I
remember how I enjoyed writing the exploit and the power that a new tool
gave me, we needed lots more tools! Oh, wow, writing this I found a
screenshot of HeapTracer showing dtlogin’s heap, I can’t believe how I
remember the shape and what each block is. Yeah, this one definitely
deserves a mention.
== CVE-2001-0550 - wu-ftpd gobbing heap overflow (arbitrary free)
Oh my god, that was a good one, the advisory even names Phrack 57! This time
is the exploit, not the vuln Not my own exploit though. I had a quite
reliable exploit, if I’m not inventing my memories, but it was
irc.segfault.net golden times, and I was there with amazing people.
That’s when I met MaXX, one of my favorite Phrack authors, who wrote on Vudu
malloc tricks, but who clearly understood free() tricks too. We then worked
together and are good friends, those were some of the best times. He showed
me his wu-ftpd exploit (or maybe he took my unfinished crappy code and
turned it into art?).
Anyway, for it to work, you had to craft a globbing pattern (*.*) so when
expanded you got an arbitrary free, and you could write what you wanted
where you wanted. You could be lazy, like me, and bruteforce the right
count ~{,,,,,,,,,...}, or you could really think about it (like he did, and
just remembered now), and figure out that if you could do more than a single
write and if you expressed the count in prime factors, you could have a
really compact globbing pattern that got expanded to overwrite really large
area. So yeah, he taught me art is a way of living, among many other things.
== CVE-1999-1085 - SSH CRC32 compensation attack
The original vulnerability (yes, I’m old, what can I remember if not old
things?). The vulnerability is that, without knowing the cryptographic
material, it’s possible to craft an ssh packet that will pass CRC32
validation in such a way that it allows a MitM to insert “keystrokes” in
the ssh stream. It’s a quite complex (at the time) cryptographic attack
that, with exploit in hand to demonstrate its power, forced all ssh
implementations, in every device and distro, to be updated with our code, as
we were starting to be known as Core SDI. Great score!
== CVE-2001-0114 - SSH CRC32 compensation attack integer overflow
Yes, almost the same name as the previous... It turned out that the final
patch for CVE-1999-1085 had an integer overflow vulnerability, which was
exploitable to get root access to any ssh server using the code. The exploit
was assigned to me (I was assigned at the time :-p ) and it turned out to be
really challenging. At some point you had to exploit the original
vulnerability, luckily I had the original exploit from 1998 by Futo and Ek,
so I didn’t have to break my head doing it. I got it working, and some time
later I found another exploit in the wild (I knew it as x1), that was very
similar (but previous apparently), and had a different solution to the
original bug. My total admiration to whoever solved the CRC32 compensation
to implement this other exploit. I remember I was really afraid we had
inserted this bug in every device, and at the same time, I wished we did it
on purpose. So I went down fishing the original email exchange to find that
it was the original ssh team, when converting from ANSI types to their type
notation, that inserted the bug, but we didn’t notice they had changed it.
Or maybe we did but didn’t say? I don’t think so, I should remember!
== BUGWEEK