The number of XCodeGhost-infected iOS apps, initially pegged at 39, has ballooned to more than 4,000.
The staggering increase was the handiwork of analysis by FireEye researchers who said that the apps were being hosted on the official Apple App Store.
"Immediately after learning of XcodeGhost, FireEye Labs identified more than 4,000 infected apps on the App Store," FireEye said.
The malicious apps steal device and user information and send stolen data to a command and control (CnC) server [and] also accept remote commands including the ability to open URLs sent by the CnC server.
These URLs can be phishing webpages for stealing credentials, or a link to an enterprise-signed malicious app that can be installed on non-jailbroken devices.
A FireEye spokesman told Vulture South that many of the infected apps were owned by "big Chinese global brands" such as consumer electronics, telcos, and banks.
The Register has asked FireEye for the names of some of the prominent affected apps and will update this story should the information come to hand.
FireEye put the huge increase in the number of discovered apps to fast scanning by its mobile threat prevention platform.
The apps were infected after developers downloaded a copy of the Xcode iOS development tool through a file-sharing service. That package was modified to trojanise apps in a way that passed App Store security checks, and was advertised on popular developer forums as a faster source to download the 3Gb Xcode file.
The success of the XCodeGhost phish is staggering; that some 4,000 apps were hosed indicates that a lot of developers were sucked into what must have been a very well-executed attack by highly capable malefactors.
Veracode principal solutions architect John Smith said that the attack challenged the notion that iOS was safer than Android.
"In recent years it has seemed that the problem of mobile malware was bigger for Android than for iOS," Smith said.
"The more rigorous testing regime required before an iOS app can be published has always been considered to be the reason for this difference, but in this case it seems to have fallen short." ®