More TFTP Honeypot Results

I’ve had some new packets show up at my TFTP Honey pot.

Previous results

Infrawatch

Beginning 2026/09/02, Infrawat.ch has sent many very mannerly, well-formed RRQ (read request) packets for a file named infrawatch-$SOMETHING.probe. The TFTP RRQ packet has 2 options, blksize:16 and tsize:0. The $SOMETHING part of the requested file name comprises 32 digits and letters. The letters are a through f. I guess these are hex digits of some sort, but I can’t see what they encode. There are 46 different filenames. So far, Infrawatch has cycled through them twice. Two packets arrive most days, separated by a few seconds. They mostly arrive between 17:12Z and 17:19Z every day. Just another company trying to scan the whole internet.

SIP phone config scanning

Sixteen IP addresses send RRQ packets for a dizzying array of file names. I think these are scans for IP-phones config files and tftpboot images. Quite a few of the file names fit the pattern [0-9][0-9][0-9]*\.cfg. Some are more obviously “phone” related, like prov/polycom/ht81x.xml or template/sip.cfg or provision/yealink/ht81x.xml

I think there are several SIP phone scanners, based on how they batch requests, and what file names they ask for.

One scanner used 5 IPv4 addresses in Contabo’s 169.58.0.0/16 CIDR. Each address asked for 40-50 different file names over the course of 3 hours. None of the 200 file names appeared in more than a single RRQ. This is pretty obviously coordinated behavior.

The IP address 45.61.185.65 allocated to “FranTech Solutions”, made 74 RRQ requests for various file names obviously related to SIP phones: cisco/spa3102.cfg, sip-ps/y000000000000.cfg, cfg/phone.cfg and the like. The 74 RRQ packets arrived within 5 seconds. Completely undisciplined access rate. FranTech Solutions is related to BuyVM of Luxembourg. One BuyVM IP address made a series of weird HTTP requests of my web server in May and June of 2026.

pxelinux.0 pugnaciousness

I also reworked my TFTP honey pot server to be less palatable to TFTP scanners. If the honey pot server receives an RRQ for a file named pxelinux.0, it sends up to 100, 512-byte TFTP DATA blocks. The first block is the first 512 bytes of a real pxelinux.0 file I found via search, the rest are randomly-chosen bytes. It waits 5 seconds for an ACK of each DATA block, which seems to be conventional.

A number of IP addresses have asked for pxelinux.0, usually in bursts of as many as my server can provide. I have a hard limit of 20 concurrent TFTP packets in combination with the 5 second timeout means my server is probably not dealing with UDP packets fast enough, and the kernel is dropping any that come in during the 5 second timeout.

I have yet to receive a single ACK or ERROR for a first DATA block. I’m not sure what the point of asking for many concurrent transfers of pxelinux.0 is. pxelinux.0 itself can’t possibly be that interesting, and a single RRQ for it would reveal if a server was offering it.