Stealth encoding bypasses IDS protection

Attack technique gets in under the wire

By John Leyden, 10th September 2001

Regcast training : Hyper-V 3.0, VM high availability and disaster recovery

Cisco's Intrusion Detection System (IDS)is not the only technology that fails to protect IIS Web servers against stealth unicode attacks.

An advisory by eEye Digital Security, reports that network and server sensors from ISS, Dragon Sensor 4.x, Snort (prior to version 1.8.1) and components of Cisco Secure IDS are affected by the issue. Symantec and Network Associates have stated that their products are not vulnerable.

Links to patches and advisories from vendors affected by the issue have been collated by Security Focus and can be found here.

Last week we reported that Cisco had to alert its customers about the problem only a day after announcing enhancements to its Secure IDS products.

In fact the non-standard method of encoding Web requests (called '%u'), which Microsoft's IIS supports but an IDS fails to decode, can allow the creation of an attack which bypasses the IDS set-ups of most vendors.

In practice, this means an attacker could modify a web-based attack, such as a "stealth" Code Red, so that requests are encoded with '%u' Unicode encoding, in order to get around IDS protection.

The obfuscation method works only because Microsoft's IIS permits a non-standard decode of html (so Apache servers, for example, are not affected).

It's worth remembering that avoiding IDS detection is only the first stage in an attack. The second stage - the compromise of the ISS Web server - is where the damage is done. Webmasters can easily stop such an attack by use of the latest security patches. But as we know, many companies are ill-disciplined in applying security patches as they come out.

IDS products, which inspect network traffic and raise alerts over suspect packets, are used for the secondary protection of IIS Web servers, so making sure they aren't fooled, still merits attention. ®