Column: The man who changed Internet security

Iria · Jul 14, 2008, 07:40 PM

While security researcher Dan Kaminsky still won't comment on the specific nature of a flaw within the Domain Name System--for fear that criminal hackers might exploit it before the worldwide network of name servers worldwide and client systems that contact them can be updated--he nonetheless went public on July 8, 2008, with some details, backed by simultaneous patch releases from Microsoft, Cisco, and others.
_______________
Source: Defense in Depth

Matth · Jul 15, 2008, 04:02 PM

Still unclear if this has anything beyond what was already known to be a problem:

1. 16 bit IDs are nowhere near big enough
2. Randomizing the source port adds another almost 16 bits
3. Some randomizers were deficient

I'm guessing that maybe some DNSes had predictable IDs and no port randomization - eg. use the DNS yourself, then attack subsequent predicted IDs, that would change the ID vulnerability from one that had to be blasted away at, to one that could be precisely attacked.

Jul 14, 2008, 07:40 PM	#1
Iria DriverHeaven Extreme Member Join Date: Apr 2004 Posts: 7,275 Rep Power: 89	Column: The man who changed Internet security While security researcher Dan Kaminsky still won't comment on the specific nature of a flaw within the Domain Name System--for fear that criminal hackers might exploit it before the worldwide network of name servers worldwide and client systems that contact them can be updated--he nonetheless went public on July 8, 2008, with some details, backed by simultaneous patch releases from Microsoft, Cisco, and others. _______________ Source: Defense in Depth

Jul 15, 2008, 04:02 PM	#2
Matth Flash Banner Hater Join Date: Jun 2002 Location: UK Posts: 3,426 Rep Power: 93 System Specs	Still unclear if this has anything beyond what was already known to be a problem: 1. 16 bit IDs are nowhere near big enough 2. Randomizing the source port adds another almost 16 bits 3. Some randomizers were deficient I'm guessing that maybe some DNSes had predictable IDs and no port randomization - eg. use the DNS yourself, then attack subsequent predicted IDs, that would change the ID vulnerability from one that had to be blasted away at, to one that could be precisely attacked. __________________ Mary had a little lamb, Her father shot it dead Now Mary takes her lamb to school, Between two crusts of bread