TO UNSUBSCRIBE: email "unsubscribe issforum" in the body of your message to [EMAIL PROTECTED] Contact [EMAIL PROTECTED] for help with any problems! ----------------------------------------------------------------------------
Not to further bash ISS, they have a good product that I depend on to held secure my enterprise. That said, they really do need to work on the signatures. The problem with false positives is wide spread throughout the IDS industry. Until an IDS maker can cut down on the false positives through better signature files, all IDS specialists are going to continue to have many headaches with this issue. I have multiple false positives daily for Glacier, Mstream_Master and Qaz_Command, if anything uses the high ports that are used in the signature it appears to set off the alert ? I'm not a programmer but it seems this could be refined a bit ? Hopefully ISS will consider an audit of there signature files soon...we can only hope they choose to lead in this area. A concerned ISS user, Chris -----Original Message----- From: [EMAIL PROTECTED] [mailto:[EMAIL PROTECTED]] Sent: Wednesday, February 13, 2002 1:36 AM To: Lau, Jamie (ISS Atlanta) Cc: [EMAIL PROTECTED]; [EMAIL PROTECTED]; Jarvis, Keith; [EMAIL PROTECTED] Subject: RE: Reading Signatures TO UNSUBSCRIBE: email "unsubscribe issforum" in the body of your message to [EMAIL PROTECTED] Contact [EMAIL PROTECTED] for help with any problems! ---------------------------------------------------------------------------- Let my just delve into my db and see how many false positives I have in there due to simplistic trojan signatures. eventname sport dport count Cisco_CR_DoS 81 7161 133 Cisco_CR_DoS HTTP 7161 26 Cisco_CR_DoS Httpd 7161 53 Glacier 81 7718 20 Glacier HTTP 7718 251 Glacier Httpd 7718 97 Glacier HTTPS 7718 9 Mstream_Master 3128 12754 27 Mstream_Master 3128 15104 43 Mstream_Master 81 12754 150 Mstream_Master 81 15104 102 Mstream_Master 81 6723 83 Mstream_Master HTTP 12754 683 Mstream_Master HTTP 15104 853 Mstream_Master HTTP 6723 1183 Mstream_Master Httpd 12754 71 Mstream_Master Httpd 15104 31 Mstream_Master Httpd 6723 67 Mstream_Master HTTPS 15104 4 Qaz_Command 81 7597 61 Qaz_Command HTTP 7597 25 Qaz_Command Httpd 7597 8 Now I know that these are false positives because in each case the source is a server that is listening on the given source port. Want some more tales of simplistic signatues going awry? We were doing some scanning the other day including some UDP port scanning. Now since the Trace_Route signature seems to look simply for the fact that a high port is being connected to I received an alert for pretty much every connection attempt. Start End Count 2002-02-11 13:01:31 2002-02-11 14:47:39 226422 That was fun. It took till around 18:00 before all the alerts were sent to the event collector. "RealSecure engine is a sophisticated state-based IDS" well lets use some of that sophistication to write better signatures. Oliver "Lau, Jamie (ISS Atlanta)" To: "'[EMAIL PROTECTED]'" <[EMAIL PROTECTED]>, "Jarvis, Keith" <[EMAIL PROTECTED]> <[EMAIL PROTECTED]> cc: [EMAIL PROTECTED], [EMAIL PROTECTED], [EMAIL PROTECTED] 13/02/2002 04:58 Subject: RE: Reading Signatures We appreciate their assistance in helping us identify the false-positive for this decode. The false positive for this decode is now understood and will be fixed in a future XPU. The RealSecure engine is a sophisticated state-based IDS that does allow us to use attributes such as directionality or session state to create accurate decodes. Occasionally, it is necessary to refine a given decode as we learn more about the reality of false-positives or false-negatives for a given implementation. Such was the case for this decode. The source and dest ports reported in any event is only intended to reflect the specific packet that triggered our event, but ultimately this can and should be compared with the documentation to interpret which address was the victim, intruder, or other. In future versions of the NetworkSensor, we plan to make improvements on how victim/intruder info is reported to the user. -----Original Message----- From: [EMAIL PROTECTED] [mailto:[EMAIL PROTECTED]] Sent: Sunday, February 10, 2002 7:40 PM To: Jarvis, Keith Cc: [EMAIL PROTECTED]; [EMAIL PROTECTED]; Lau, Jamie (ISS Atlanta); [EMAIL PROTECTED] Subject: Re: Reading Signatures > I wrote the Asylum backdoor signature (and most backdoor signatures in > RealSecure, if memory serves me) and this is the behavior of many of them. So you're the one to blame. > The reason the source and destination ports are apparently swapped is > caused both by how the signature actually detects the backdoor traffic and > how the event is reported within the network sensor. Lets try to get something clear first. Is Realsecure reporting the source and destination ports relative to the connnection or relative to the direction of individual packets? This is a very important point. If Realsecure is trying to display the direction of the connection, then src and dest ports are not "apparently swapped", they ARE swapped. > Many of these signatures operate by searching through TCP and UDP traffic > for so called "banners" announced by backdoor and trojan software. Naturally, > this traffic originates from the infected computer and not from the attackers > computer. No. The traffic is originating from the attacker. The attacker is initiating the connection to the trojan port. It is true that the PACKETS carrying the banner will originate from the trojan port, but what is important is the direction of the connection not individual packets. > This explains why you will see an event, for say "Asylum," which > will show an originating TCP port of 23432 and not the ephemeral port most > likely doled out by the attacker's OS. Well there's a problem then. Not with documentation as you say below, but with the fact that the signature is too simplistic. Simply searching for "banner" and packet source port = 23432 is not enough, the signature needs to keep track of the direction of the connection. ie search for "banner" and packet source port = 23432 AND connection dest port is 23432. > Ultimately what this means is that the RealSecure help sort of makes sense > once you know the above details on how these signatures work. However, it > certainly could be improved to describe the operation of these signatures > in greater detail. I will talk with our documentation group about having > these blurbs improved in a future release of RealSecure. Having help that only "sort of" makes sense AFTER extra info is ellicited is hardly good enough. > I previously replied to Mr. Wangler about his inquiry, but unfortunately > Cc'd the incorrect address for ISS-Forum. The traffic he was seeing trigger > this event was in fact a false positive, caused by an FTP 'PASV' command > originating from an ephemeral which just happened to be 23432. This > particular signature searches for a banner of "PAS", among others, which > saw "PASV" and thought it was seeing backdoor traffic. This false positive > will be corrected in XPU 4.1, due out shortly. This false positive wouldn't occur if the direction of the traffic were tracked. At least one positive thing came out of this, the fact that there is more to the signatures than simply the ports used. So from what you've said, Realsecure is little more than a management interface onto of ngrep, or perhaps "snoop | grep". I probably shouldn't cloud the issue, but on the topic of simplistic signatures, how about these URLs being considered as HTTP_Shells /shareholdercentre/ /bb/investorCentre/shareholderMeetings.jsp /bb/investorCentre/shares.jsp my guess would be that the signature simply reads: alert if URL contains "/sh" This speaks volumes about Realsecure as a product. Oliver "Keith R. Jarvis" To: Oliver Friesen/AUST/CSC@CSC <[EMAIL PROTECTED] cc: [EMAIL PROTECTED], [EMAIL PROTECTED], [EMAIL PROTECTED] > Subject: Re: Reading Signatures Sent by: owner-issforum@i ss.net 09/02/2002 03:23 TO UNSUBSCRIBE: email "unsubscribe issforum" in the body of your message to [EMAIL PROTECTED] Contact [EMAIL PROTECTED] for help with any problems! ---------------------------------------------------------------------------- I wrote the Asylum backdoor signature (and most backdoor signatures in RealSecure, if memory serves me) and this is the behavior of many of them. The reason the source and destination ports are apparently swapped is caused both by how the signature actually detects the backdoor traffic and how the event is reported within the network sensor. Many of these signatures operate by searching through TCP and UDP traffic for so called "banners" announced by backdoor and trojan software. Naturally, this traffic originates from the infected computer and not from the attackers computer. This explains why you will see an event, for say "Asylum," which will show an originating TCP port of 23432 and not the ephemeral port most likely doled out by the attacker's OS. The cause is not related to load balancing, or asymetric routing, on the monitored networks. This may affect RealSecure's ability to reassemble TCP data streams but each packet will still maintain its original 'dest' and 'src' port data fields. Ultimately what this means is that the RealSecure help sort of makes sense once you know the above details on how these signatures work. However, it certainly could be improved to describe the operation of these signatures in greater detail. I will talk with our documentation group about having these blurbs improved in a future release of RealSecure. I previously replied to Mr. Wangler about his inquiry, but unfortunately Cc'd the incorrect address for ISS-Forum. The traffic he was seeing trigger this event was in fact a false positive, caused by an FTP 'PASV' command originating from an ephemeral which just happened to be 23432. This particular signature searches for a banner of "PAS", among others, which saw "PASV" and thought it was seeing backdoor traffic. This false positive will be corrected in XPU 4.1, due out shortly. Hope this helps, - --krj > > I think that's what I said. > > However I don't think it is specific to the signature, but rather, as I > said, because Realsecure is not seeing the traffic in the correct > direction. I would guess that this is not a fault with Realsecure but > rather a consequence of load balancing in the network, though I haven't > verified this with a sniffer. > > Oliver > > > > > "Lau, Jamie (ISS > Atlanta)" To: "'[EMAIL PROTECTED]'" <[EMAIL PROTECTED]>, Dan Wangler <[EMAIL PROTECTED]> > <[EMAIL PROTECTED]> cc: > Subject: RE: Reading Signatures > 08/02/2002 06:58 > > > > > > > This signature will show source and destination reversed. Several of the > backdoor signatures do this. > > -----Original Message----- > From: [EMAIL PROTECTED] [mailto:[EMAIL PROTECTED]] > Sent: Wednesday, February 06, 2002 6:35 PM > To: Dan Wangler > Cc: [EMAIL PROTECTED]; [EMAIL PROTECTED] > Subject: Re: Reading Signatures > > > > TO UNSUBSCRIBE: email "unsubscribe issforum" in the body of your message to > [EMAIL PROTECTED] Contact [EMAIL PROTECTED] for help with any > problems! > ---------------------------------------------------------------------------- > > > If the events you are presenting are partly correct, then I would say that > you have reversed the source and destination ports because you know that > the traffic was was not going from source port 21 to destination port > 23432. But in a load balanced environment the IDS does sometimes see > connections in the wrong direction, and hence often comes up with false > positives for trojan port activity. > > Oliver > > > > > > Dan Wangler > > <[EMAIL PROTECTED] To: [EMAIL PROTECTED] > > > cc: > > Sent by: Subject: Reading Signatures > > owner-issforum@i > > ss.net > > > > > > 07/02/2002 08:51 > > Please respond > > to Dan Wangler > > > > > > > > > > Maybe someone can help me here. > > I seem to be misinterpreting the events I have been collecting. For > instance, > yesterday, I saw 7 hits of the Asylum signature. One of our network > sensors > reported the following information (I have sanitized the information > because I > have not determined this to be a true event!): > > Alert Type: Suspicious TCP > AlertPriority: 1 > AlertID: xxx... > Source IP Address Name: (One or our DMZ servers) > Destination IPAddress Name: (An IP Address from a block owned by a > Florida > ISP) > Source Ethernet Address: (Intentionally omitted) > Destination Ethernet Address: (Intentionally omitted) > Source Port: 23432 > Source Port Name: 23432 > Destination Port: 21 > Destination Port Name: FTP > Protocol Id: TCP(6) > Data: PASV > > The explanation for Asylum (from the online RealSecure Help documentation) > is > as follows: > Asylum RAT (Remote Access Tool) backdoor for Windows (Asylum) > About this signature or vulnerability > This signature detects a TCP connection to an Asylum backdoor on port 23432 > on > your network. > Default risk level > High > Sensors that have this signature > RealSecure Network Sensor: XPU 2.1 > Systems affected > Windows NT, Windows 95, Windows 98, Windows 2000 > Type > Unauthorized Access Attempt > Vulnerability description > The Asylum remote administration tool (RAT) is one of many backdoor > programs > that attackers can use to access your computer system without your > knowledge or > consent. With the Asylum backdoor, an attacker can upload and execute files > on > the host and restart the computer. > Asylum is distributed with an "edit server" program that allows the > attacker to > customize the backdoor server to run on arbitrary ports (TCP 23432 by > default) > and use combinations of startup methods, making it difficult to remove from > an > infected host. > (I have intentionally omitted the How to Remove section of the Help > document.) > > In the description of the signature, it says "This signature detects a TCP > connection to an Asylum backdoor on port 23432 on your network." However, > the > source address of this attach is port 23432 not the destination. I have > interpreted statements on other signatures that state "... connection to > ..." > to mean connecting from a source address to a destination address. Is this > not > the way I should be interpreting it? This particular source machine is in > our > DMZ, and if it were attacked from the internet, I would hope our external > sensor would have detected an event with a source IP to be something and > the > destination IP address to be 23432. Our external sensor report no event of > anything trying to connect to port 23432 in our network. > > Am I misinterpreting the signature explanations? > > Thanks > > > Dan Wangler, GCIA, IT Security Administrator > IT Security Response Team, Texas Instruments, Inc. > Spring Creek Bldg 1, C196, > 6500 Chase Oaks, Blvd, MS 8417, Plano, Texas, 75023 > Tel #; 214-567-8304; Email:; [EMAIL PROTECTED] > > > > > > > > -- Keith R. Jarvis ([EMAIL PROTECTED]) http://xforce.iss.net Senior Attack Detection Logic Systems Design Scientist, X-Force Internet Security Systems, Inc. +1-404-236-2833 (direct) The Power to Protect +1-404-236-2624 (fax)
