Internet attacks are a form of a public health problem, and this months Top 20 list of Internet security risks offers tactics to contain this costly and potentially contagious nuisance.
Like the cold-causing rhinovirus, most Internet attacks are spread through routine social contact (albeit electronic, rather than physical). Like Internet attacks, conversely, the rhinovirus comes in hundreds of varieties and has no cure. The only effective counterattack is a change of personal habits.
The new Top 20 list says, in effect, "Users, wash your hands!"
The formerly emphasized Top 10 vulnerabilities were identified in June of last year by the SANS Institute with input from the CERT Coordination Center, The MITRE Corp., the National Security Agency, the Department of Defense and many technology vendors. The risks on that list were serious defects in widely used pieces of Internet software and often surprisingly old; their remediation was the purview of system administrators: for example, loopholes in the BIND (Berkeley Internet Name Domain) implementation of the Internet Domain Name System because BIND is so widely used and because its loopholes enabled a wide range of distributed attacks.
Sendmail exploits and buffer overflow exposures were other representative entries on the Top 10 list; by contrast, user-level loopholes such as poor password practices didnt appear until No. 8 on the list.
The Top 20 list, jointly released by SANS Institute and the FBIs National Infrastructure Protection Center, reverses that order of emphasis. Poor password practices have vaulted all the way up to No. 2 in the list of General Vulnerabilities, bracketed at No. 1 and No. 3 by two risks not even mentioned before: default installations of operating systems and applications at No. 1 and inattention to backups at No. 3.
Default installations create several risks. Absolute path names to critical files, such as e-mail address books, can readily be predicted since most users accept the default path names offered by installation utilities.
Additional exposure comes from collections of sample code, furnished to demonstrate software features and written for clarity rather than robustness. An attacker, knowing that these samples may be present, can exploit them—with the legitimate users privileges—by any of several well-known strategies.
Most dangerous are the many services, often network-accessible, enabled by default during installation of products such as database servers. This proved to be the Achilles heel that exposed the database in eWeeks Openhack challenge during June of last year, when the default password for the mdsys administrative account was accidentally left unchanged. Like several other Oracle Corp. server default accounts, the mdsys default password is especially easy to guess—its the same as the user name for that default account.
It may seem like conceding defeat to list inadequate backups as a security vulnerability, but eWeek Labs sees this as a welcome focus on solving the real problem—loss of productive time—rather than treating security as a futile exercise of fixing an engine while its running at top speed. If the damage done by Internet attacks is corrected quickly and inexpensively, the appeal of such mischief will rapidly fade.
Likewise, improved logging (No. 6 on the list) represents pragmatic acceptance that no system is perfect. Security is not just prevention—its also detection and response. Learning from inevitable failure is essential to long-term success.