HARING PRACTICAL PROTECTION IT SECURITY MAGAZINE Nelo) ag ARE YOU GETTING THE,MOST OUT OF YOUR IPS? [ans Peta ea te yet Peta ene NOTES OF THE NETWORK ADMINISTRATOR IMPROVING YOUR CUSTOM SNORT RULESHARINS IT SECURITY MAGAZINE, PRACTICAL PROTECTION HARINS =" Editor in Chit: Amat Letner mata lotr sotware. comp! ‘olin lesinskod sofware com pl Editorial Advisory Board: Rabacea Wyn, Michal Munt DTP: ronausz Pogroszowsi [An Draco: mses Porm ones pograzenstontwre. can pl Proofreaders: Bary McClain, Mark Lohman, Graham Hit Top Batatestrs: Rebecca iy, Bob Fozen, Cars Ayal, Stave (Goce Maton Dumas, Andy Aven ‘Special Thanks othe daa testers and Proofreaders who Papo Unt te ese, Winout ther aetancs tere wouls no be akin magazine CEO: Eva Lozowoka (na lozowickag@sotar compl Production rector: Ane Kus antag hight ony Jalna lesinska@hakins org ‘Subscription: ona Brzoric ana beri @olare com pt Publisher: Sofware Prass Sp. 2.0.9. Phone: 9173382631 vwwnrakind orgfon ‘Wat every effort has baen made to ensure he high aualty of the maga, the odors make no waren, expres or impl, [A vade mara presented inthe magazine were used only for Informative purposes. Alright trade marks presaradn the magazine are reserved by tbe companies which own them ‘Tocrete graphs and dagrama we Used saxty. prog by ee Soar “The ators use automate OTP sya ACCIMCIS Natnematcal mules crested uy Design Solonce Math Type™ DISCLAIMER! The techniques described in our articles may only bbe used in private, local networks. The editors hold ne responsibility for misuse of the presented techniques or consequent data loss. AH ARING scarterkit Dear Readers, As you already know Snort is the most widely deployed IDS/IPS technology worldwide. Developed by Sourcefire, Snort combines the benefits of signature, protocol, and anomaly — based inspection. In Snort Special Issue Leon Ward, Joel Elser, Kishin Fatnani, Shivang Bhagat and Rishita Anubhai provide insight into writing Snort rules and into deployment of this IDS/PS. With the end of the year inevitably approaching, it's high time to briefly reflect on 2010 and enter 2011 with new solutions and ideas for the foreseeable future. ‘Some of them are provided by KK Mookhey in "How to get the most out of your IPS?” And annual Conference on Nagios and OSS Monitoring is to be looked forward too. Wishing you wonderful Christmas, Haking Team TOOLS 4 Uptime IT Systems Management Review by Michael Munt 6 Notes of the Network Administrator by Doug Chick | recently used SNORT and another program | like EtherApe to detect a major intrusion on my network. Within minutes millions of people were on my private fiber network. Once I isolated the problem | immediately connected my Internet provider. Like with many ISPs they denied it and recommended | look at my routing tables, Ifyou are a network manager then you know in very many cases you must provide proof to your ISP before they are willing to provide you with support. In this case | recorded the event showing that there was hundreds of thousands, perhaps even a milion people was passing traffic on my network. I sent the logs, and a video ‘of my SNORT and EtherApe displays and emailed them to the ISP. | then shutdown the two interfaces on my router and waited for a return call. The call came quickly to. wewchaking org8 Writing Snort Rules by Kishin Fatnani ‘Though Snort can also be used for packet logging, sniffing or as an IPS, however in this article we wil look more into the concept of rules by which Snort detects interesting trafic ‘or us, basically the kind of traffic we are looking for, ike @ network attack, a policy violation or may be traffic from a network application or device that you are troubleshooting 14 Collection and Exploration of Large Data by Luca Deri Collecting and exploring monitoring data is becoming increasingly challenging as networks become larger and faster. Solutions based on both SQL-databases and specialized binary formats do not scale well as the amount ‘of monitoring information increases. In this article 1 would like to approach to the problem by using a bitmap database that allows to implementation of an efficient solution for both data collection and retrieval. NetFlow and sFlow are the current standards for building trafic monitoring applications. Both are based on the concept of a tratfic probe (or agent In the sFlow parlance) that analyses network traffic and, produces statistics, known as flows, which are delivered to a central data collector. As the number of flows can be pretty extremely high, both standards use sampling mechanisms in order to reduce the workload on both of the probe and collectors. ADVANCED 48 Improving your custom Snort rules by Leon Ward Wile itis easy to create a custom Snort rue, do you know if you are actually making @ good one or not? This atte introduces some common mistakes | find in custom Snort rules and the potential implications of those mistakes. The Snort IPS engine has changed substantaly over the last ten years. Packet processing speed has improved, IP defragmentation and. stream reassembly functions have evolved, the connection and state tracking engine has matured, but there is one thing that keeps getting lft behind, Custom rule-sets, With each revision of Snort new features are added that enhance the detection capabilty and aid in packet processing performance of the Snort engin. 24 An Unsupervised IDS False Alarm Reduction System —- SMART by Gina Tihai and Maria Papadaki Signature-based (or rule-based) network IDSs are widely used in many organisations to detect known attacks (Dubrawsky, 2009). A common misconception about IDSs is that they are, Plug-and-Play devices that can be installed and then allowed ‘to run autonomously. In reality, this is far from the truth, wwrwhaking.org ee ANERY 30 Content Modifiers: Keep it Specific by Joel Esler Witnout going off the deep-end here and discussing every single Snort rule keyword, | just wanted to touch on a fow modifiers that people sometimes misunderstand, These modifiers are not keywords of themselves, but rather they apply as modifiers to another keyword. That Keyword is content, The content keyword is one ofthe easiest pieces of the Snort rules language as all it does is look fora particular string, The modifiers that | am talking about are: 1. Offset, 2. Depth, 3. Distance, 4. Within, 5. nocase, 6. hit 7. rawbytes. DEFENSE 34 Deploying Snort as WAF (Web Application Firewall) by Shivang Bhagat and Rishita Anubhai In today’s environment, web applications are becoming a popular attack point with attack agents. Atack agent can bbe a human altacker or an automated worm. It tries to explore vulnerabilities over HTTP(S) and exploit it for a given. ‘opportunity. The web application landscape is also changing ‘and more complexities. are goting added, it provides ‘openings for vulnerabilities and possible expotations. HTTP trac is no longer restricted to name-value pais and traditional HTML only. thas evolved with Web 2.0 and RIA, it allows JSON, AMF, XML and various other structures, It has become a platform for robust and advanced business application hosting and usage. It is imperative to secure business applications against all possible attack vectors and to maintain security of information and access. In this, aticle we wil try to understand Sor from HTTP standpoint ‘and how we can protect applications for some of the popular attack vectors lke XSS or SOL injections by using It 40 Are you getting the most out of your IPS? by K. K Mookhey Picture this: @ muli-ilion dollar global telecom giant has invested millions of dollars into building a state-of-the-art Security Operations Center. They have @ huge display screen being monitored 24/7 by a team of specialists who = 50 we are fold — have received extensive training in the specific technologies used, as well as in the overall incident management framework. They've deployed a high-end intrusion prevention system (\PS) which feeds into their ‘Security incident Management (SIM) system. A review of the procedures and Service Level Agreement (SLA) of the ‘SOC team signed wit the rest ofthe business reveals that they are ready to respond 24/7 and have committed that Within 2 hours of @ serious attack they will respond to any serious attacks. On paper it all looks impressive and too 00d to be true. SNORT HARING I:TOOLS up.time IT Systems Management Review When it comes to the performance and availability of your IT infrastructure and applications, deep, and easy-to-use monitoring is a must. than ever, as applications and services span many environments (cloud, virtual and physical) and infrastructures (Windows, UNIX, Linux, VMware, etc) ‘Additionally, IT infrastructure is now global and monitoring from one tool, instead of many point tools, is essential to drive down costs while increasing performance. up.time’s Single pane of Glass dashboard provides a deep, easy-to-use, affordable and complete IT systems management and monitoring solution designed for mid-enterprise companies. Every license in up.time’s comprehensive, cross platform management and ‘monitoring suite includes unlimited access to: Ts monitoring and reporting is more complex + Server Monitoring + Virtual Server Monitoring + Cloud Monitoring + Co-Location Monitoring + Network Monitoring + SLA Monitoring & Management + Virtualization & Consolidation + Capacity Planning + Application Monitoring + Application Transaction Monitoring + Proactive Outage Avoidance + IT Process Automation (One of the highly beneficial capabilities of the uptime suite is access to Service Level Management. Most departments require SLA’s (Service Level Agreement) for their equipment and applications. uptime makes it very easy to define and document agreed SLAs, then link them through to the appropriate infrastructure service. up.time also provides the ability to automate responses to issues, removing the possibility of human error while greatly decreasing the Mean-Time-To- Repair. In fact, uptime goes a step further and lets HARING SNORT SLA for ou uptime SLA for Payroll SLA for SLA for Email ERP SLA for CRM $395 per Windows Server $695 per UNIX Server $695 per ESX Server (no charge per instance or VM) Alkin-One: No additional charges for modules or applications. UL: http://wwwuptimesoftware.com/ administrators proactively automate responses based ‘on thresholds, allowing uptime to solve problems before they happen. It's not just physical, virtual and cloud based servers that up.ime monitors, it also provides application and application transaction monitoring across Email, CRM, ERP, Web, and even custom applications (including any 3rd party commercial software or in-house developed applications). In addition to all of the above, up.time is extremely advance with its reporting. This area is a major asset of the up.time suite and it is immediately apparent that wewchaking orgup.time IT Systems Management Review the reporting has had a good deal of thought and time spent on it. The reporting is both deep and very easy to create. Administrators can generate and save reports in different formats and quickly send (or set automated daily, weekly or monthly sends) via email to a single user or an entire group. When we say up.time is easy to use, we really mean it Installation of up.time was a dream, very simple, straightforward and easy to do. The entire process only takes a few mouse clicks. If you decide to go with the VMware appliance option, this is even easier as it comes as a pre-installed appliance that is can be imported into any virtual infrastructure. Managing the monitored environment is achieved through a web portal which is simple, clean and easy to read (unlike other monitoring solutions that appear to have far, far too many menu's and options). Once you enter a small amount of information, the ‘My Portal home page is displayed. This page provides a summary list of the current alerts that you have configured together with saved reports and support links. All the tutorials are web based and, again, very clean and concise. The end result is that you are up and running with this product very quickly. Everything about the product screams simplicity and yet it's extremely deep in its monitoring and reporting capabilities. Compared to other tools, up.time is very refreshing. It's certainly powerful enough for large enterprises to monitor over 5,000 servers and applications globally, and yet its affordable for the small and mid enterprise companies to monitor between 25 and 500 servers and applications. The help documentation is included and available through your browser locally on the server. uptime uses one of the friendliest licensing models in the industry, with its per-physical-server licensing across the board, even in virtual environments. All you need to do is count the number of physical servers you want to monitor, and that’s it. Everything is included, no modules, hidden extras, application charges or management packs. There is so much depth to this product that | can’t comment on it all within the scope of this article. If this sounds interesting, up.time makes trying it for yourself incredibly easy. | suggest downloading the trial and taking it for a test drive. | think you'll be as impressed as | was. In fact, this product is so good, I'm starting to recommend that my clients review their monitoring needs and consider trialing in up.time. MICHAEL MUNT up times Complete IT Infrastructure and Application Monitoring that’s DOA ee (© Deep: Granular Monitoring and Reporting ac (VMware, UNIX, Windows, Linux, Novell and more) and locations. oar Cees Eee minutes or less. No consult Cee ence) Ca Per-Physical-Server Pricing, even in virtual environments. Poses ta ne too has eee ee er ree es ee) Ge eer aes eae enue la ro Pn Syren Cen? Solve your Systems Management needs at www.uptimesoftware.com ® Learn: ® Watch: ® Evaluate: White Papers & Product Tour 30-Day Risk-Free up-time Case Studies ‘and Videos Enterprise Trial cause downline is nat an antonBASICS Doug Chick printable: Notes of the Network Administrator | have computer networking friends that work with various departments of the government, corporations and private companies that are very aware of the possible threats to their computers and networks. What you will learn... overview of 10S + overview of Snort What you should know... + basic knowledge of TCP/IP hey are and have been taking serious steps to | secure their systems, despite litle interest or concern fram company or agency managers. Because of this lack of concern, many network security, managers must take it upon themselves to secure their networks, but with litle or no budget, they must rely or Open Source software to do it One such software | use is SNORT. SNORT is an open source network intrusion prevention and detection program (NIDS). This is a must have in any network managers security toolbox. If you are a Linux fan, then I'm sure you already know about SNORT, as it comes preinstalled with such Linux installs as Back/Track 4 and Knoppix-STD. Note there is also a Windows install SNORT, monitors, captures and analysis incoming packages and scans for a pattem of intrusion. In other words, it looks for specific packet signatures used by hackers, and automated hacking programs. Snort can detect attacks, probes, operating system fingerprinting, buffer overflows, port scans, and server message block scans. (In other words; all incoming network traffic.) | recently used SNORT and another program | like EtherApe to detect a major intrusion on my network. Within minutes millions of people were on my private fiber network. Once | isolated the problem | immediately, connected my Intemet provider. Like with many ISPs they denied it and recommended I look at my routing tables. If you are a network manager then you know in very many cases you must provide proof to your ISP before HARING SNORT thoy are willing to provide you with support. In this case | recorded the event showing that there was hundreds of thousands, perhaps even a million people was passing traffic on my network. | sent the logs, and a video of my SNORT and EtherApe displays and emailed them to the ISP. | then shutdown the two interfaces on my router and waited for a retum call. The call came quickly too. The ISP's main core router was hacked, and th routes were re-directed. Two hours later all the ISPs network engineers were called in. | stopped it on my end by shutting down the two interfaces it was coming in from, but it took them two more days to correct it ‘on their end. | have redundant circuits from another provider, so | simply used those. The direct impact to me was minimal. Stil, with the flood of hundreds of Figure 1 wewchaking orgthousands of people directed to my private network. with over twenty offices connected, | am still waiting to discover any long term damage. Privately one of the techs from my ISP told me later that they thought the intrusion came from China, With the help of SNORT and EtherApe | was immediately alerted to a flood of unwanted traffic to my network. To me this reaffirmed the necessity of intrusion detection programs, and also made me research how many more types there are. Types of Intrusion Detection Systems: Intrusion prevention systems (ISP) This device, also know as Intrusion Detection and Prevention Systems are network security appliances that monitor for malicious activity. This is a stand alone appliance identifies suspicious activity, isolates and logs it, and attempts to block Host-based intrusion detection systems (HIDS) Host-based intrusion detection systems are installed on the computer level and monitor the actual server it is installed for suspicious activity, where ISPs operate on the network and analyze packets. Protocol-based intrusion detection systems (PIDS) This detection system is generally added in front of a web server and analyzes the HTTP, (and HTTPS) protocol stream and or port numbers. Application protocol-based intrusion detection systems (APIDS) APIDS typically are placed between servers and monitors the application state, or more accurately the protocols being passed between them. For example a web server that might call on a database to populate a webpage field. like SNORT because; one itis free, and two because of the support and its sheer flexibility. ( know, that is three things) SNORT RULES. Like with any intrusion detection device, SNORT has Rules. (concent: /00 01 86 a5/") meg:"nouncd ac The rules are actions that tells Snort what to do when it finds a packet that matches the rule criteria. There are 5 default actions; alert, log, pass, activate, and dynamic. If you are running Snort inline mode, there are additional options which include; drop, reject, and sdrop. wwrwhaking.org alert — generate an alert using the selected alert method, and then log the packet og — log the packet ass — ignore the packet 4, activate ~ alert and then turn on another dynamic rule 5. dynamic — remain idle until activated by an activate rule , then act as a log rule 6. drop — block and log the packet 7. reject — block the packet, log it, and then send a TCP reset if the protocol is TCP or an ICMP port unreachable message if the protocol is UDP. 8. sdrop — block the packet but do not log it. If you want to learn more about SNORT, | recommend you visit there site at: www.snort.org | know there are other NIDS programs out there, and I'm sure they are just as good as SNORT, but as a network administrator/engineer this particular program has already proven itself to me. EtherApe: Gorilla Virtual Warfare As | mentioned before, another program | like is EtherApe. EtherApe is a graphical network monitor for UNIX modeled operating systems. It doesn't have the same features as SNORT, but what is does do is sives you a graphical overview, on what is going on in ‘your network. I run EtherApe on a large screen monitor above my desk. When trouble comes, | can see an immediate flash of color that warns me that there is a possible situation on my network. This seemingly simple program has called me to action a couple of limes. EtherApe has the ability to filter just the port number you want to monitor, or by default all of them. Like the name implies it works on an Ethemet network, but it also works with FODI, Token Ring, ISDN, PPP. and SLIP. DOUGLAS CHICK Douglas Chick is a Director of information Systems for @ large ‘company in the Orlando FL area. Although he introduces himself as a Network Engineer/Administrator. As with many ‘computer people, Doug holds an MCSE and CCNA certification. ‘Doug frst became known en the Internet in May of 2000 for @ series of articles about Microsoft retiring the NTA MCSE that were published on over 30 Internet Magazines. With his humor and insightful look into the world on computer professionals, he receives a great deal of attention from other computer professionals around the world. Andis proud to admit that less that one percent are in response for his many typo's and mis- spellings. For more visit: www.TheNetworkAdministrator.com SNORT HARINYI,BASICS Writing Snort Rules Snort, as you would know, is a tool used to detect intrusions ona network. What you will learn... + ew get you started with wring basic Snort rules + Configuration of Snortis not covered here + Preprocessor and advance rules are not covered What you should know... + Good knowledge about TCRIP networks «Packet Analysis + Using and configuring Snort logging, sniffing or as an IPS, however in this article we will look more into the concept of rules by which Snort detects interesting trafic for us, basically the kind of traffic we are looking for, like a network. attack, a policy violation or may be traffic from a network. application or device that you are troubleshooting. For instance, if someone is doing XMAS port scan to our network using nmap with the -=x option, Snort will give Us the following alert message. Ton the tool can also be used for packet nex ge ont Attempted Information Leak) [Priority: 2) 15351246. 970825 192.166.0.11 192.168.0,1:132 ‘pep "7:53 205:0x0 zp:28021 rpuani20 ogmuan:40 CABTOFBIES Ack: ¢x0 Win: OxBOO TepLen: 20 Urgetes ox0 iret => /fewu.omergingtneeats.nat /egi-bin/cvsweb ogi! sge/SCRN/SCRO NAP] [Reet => neepe// doc semergingthreats.net/2000566 Ifthe use of P2P or IM applications is against the corporate policy, Snort can detect their use on the network and provide alerts with messages similar to these: [e+] Pap Bivtortent announce request” (+7) or H@AHING sNorT [99] CHAT Yahoo IM successful Logon [*1 To identify and alert on the appropriate events of interest, the detection engine needs to have rules for each event, A rule is what tells the engine where and what to look for in the network traffic and what needs lo be done if detected. Here is a simple rule and the alert generated when that rule is triggered. Rule: fleet tep any any -> 292.168.9.1 40604 InegstAcceas to port 40804 on server"? sid:3000001;) Alert: [#5] [1:3000001:0} Access to port 40404 on server [*4] (priority: 0) TCP TTL:64 TOS:0K0 1926265 IpLen:20 Denlen: 40 . st OXSISTTI0 Ack! OXESESTPY Wn: 200 tepten: 20 Looking at this you must have got some idea about the components of a rule, but we shall go in to further depth and look at various options that can be used in the rules. So let us start with the above rule which is @ quite simple one. Arule is divided into two parts: + Rule Header — the initial portion before the parentheses wewchaking org+ Rule Options — between the parentheses The Rule Header ‘The header tells the engine about which packets to look into based on the protocol, IP & Port addresses and the direction in which the packet is travelling. The action to be taken upon detecting an event is also mentioned in the header. The following is the rule header format: Let's take a look at the first field in the rule header which is - This field shows the action that will be taken when @ packet matches the rule. Typically you would see alert or log here, which are self explanatory, but there are some more actions like pass to ignore the packet or activate to invoke another rule which has been idle due to its dynamic action. Snort will match the addresses and other options only if the packet is part of the protocol mentioned here, Which could be 1», tem, cep OF udp, IPA R For the rule to be triggered, the IP address specified here must match the IP address in the packet. Each packet has an IP header which has two IP addresses, source and destination. It will depend on the direction field whether the IP address given here will be matched with the source or the destination. This field can have a specific IP, network address with CIDR notation, multiple IP addresses enclosed within square brackets and separated by comma or the word ‘any’ to ignore the IP address. We can use variable representing IPs of servers or networks which makes it easier to manage. The variable are declared initially with the keyword va: and subsequently the variable name preceded by a dollar sign can be used in the rules. Variable decalaration war TEST_SERVER 192.168.0 Using the variable cp any any > wwrwhaking.org There are some predefined variable names for IPs and networks which are used in the default ruleset, some of them include If the protocol used is TCP or UDP, there will be a respective header attached to the packet which will have 2 port addresses, source and destination. Again, for the rule to be triggered, the port nos. must also match. Whether this field will be matched with the source or destination port will depend on the direction field, A port number may be a single number or a range given with the - separating the lower and upper boundaries e.g. 1:200 for port numbers from 1 to 200, Justas in IP ADDR, variables can be used to represent, port numbers of services like this: Variable declaration Using the Variable ‘There are some predefined variable names for ports which are used in the default ruleset, some of them include — : This field stands for the direction in which the packet must be travelling to match this rule. If the direction given is -> then the IP and port addresses on the left side of this sign will be matched with the source IP and port while the right side will be matched with the destination. In case of the < sign, the left side address will be checked in source as well as destination and if found in one, the right side will be matched with the other. Note The exclamation sign : can be used to negate the IP ADDR and PORT NO. fields Coming back to our previous example, 80904 on server" si we are telling Snort to alert us if it sees a packet with any IP and any port number in the source (being left side of the direction -> field) and the IP SNORT HARING |.BASICS address 192.168.0.1 and port number 40404 in the destination. Now suppose there is a service running on port 40404 on the server 192.168.0.1 and some client connected to the service and transmitted and received several messages. If this communication is captured by Snort, we surely are going to receive the alert as given above. The question here is that will Snort alert us while the connection is being established, after it has been successfully established, at the first message, or at all the messages sent and received?? To know that we need to look at the rule and its description below it which clearly says that an alert to be generated whenever any packet is seen to be going to t service. By this explanation itis very clear that we are going to receive a lot of alerts as Snort will be looking at the communication packet by packet and each packet going towards the server will trigger the alert, Hence, we are going to get an alert right from when the client requests connection to the service (SYN packet), then the acknowledgement from the client to server, each packet containing data (message) or acknowledgement from the client to server, the finish (FIN) and finish acknowledgement packets. For any packets from the server to the client, there will be no alert, hence the SYN/ACK will not give an alert, neither any messages received by the client. So, does that serve the purpose? Well, it depends on what was the objective for writing such a rule. If you are just looking for attempts to connect to the service or successful connections, then this would create a lot of noise by giving too many alerts which are not required. However, it may be useful if you are trying to debug or troubleshoot a network application or analyze malware activities like if you want to see how frequently is the application sending a keep-alive message while being idle or if you want to check on the activity of a bot To reduce the noise and get a single alert on connection or when a specific message is being sent, we need to use the rule options which are enclosed within the parentheses. The Rule Options We have already used a couple of options in our sample rule but those options did not have any role in the detection logic. These options are called general or metadata options which provide information about the rule or event and do not have affect during detection The description of the event which is displayed in the alert comes from the rule option ms. This option consists of the keyword ==: followed by an argument, the message itself, separated by a .. Each option is then followed by a ; to end the option and begin another, if any. Some options do not require an argument hence vIHARING snort they also don't require the is still required . however, the - sid and rev ‘The other option that we have in our rule is the ‘sid’ option which uniquely identifies Snort rules, Each rule in Snort has a unique sid which is provided by the author of the rule. To ensure that it does not conflict with the sid of the rules included in the Snort distribution, the sid must be above 1000000. Other community rules may have their own series which must be avoided in custom rules. The sid is often accompanied by the send (IP dat="192. 168,021") /RePeport=404040) Making the rule more specific ‘We do not want alerts for each and every packet of the connection as just the knowledge about the connection attempt would be good enough. In this case we can add another option which will make the rule trigger only at the connection request from the client side. As we know that a connection request packet has only the SYN bit sel, so we can instruct Snort to look for this bit. The ‘option for this is c1ass and it takes the initial letter of the required flags as the argument ike the one here” alec: ‘ep any any => 192.166.0 30000017) wychaking orgHere we are saying that alert us only ifthe TCP header in, the packet has the SYN bit exclusively set. This will add to all the previous criteria of Protocol, IPs and Ports as specified in the rule header. Now we will not get all the noise we were getting earlier but there is stil a problem which can give us false negatives i.e. it may not alert us even if the connection is being requested on the said service. This is because the first packet in the three-way handshake in TCP can have an ECN flag set along with the SYN flag. The two ECN flags were reserved earlier, hence they are represented by the number 1 and 2 rather than the initial alphabet. We now need to alter the flags arguments to account for the possibilty of an ECN flag being set. This can be done by adding the optional flags preceded by a comma, as given below: “access t0 port on server’; idi30000 The flags option affects the detection, hence it cannot be called a general or metadata option. The options affecting detection are payload or non-payload options. As the flags option refers to the information in the TCP header and not the payload, hence it isa non-payload option. Payload Options Most rules do not just depend on the header fields to identify an event of interest but they look into the payload i.e. the data passed in the packets. The headers are based on standard specifications, hence Snort can identify each field of a header and lets us refer to them by keywords (e.g. flags, tt, id), however, the payload may have variable data depending on the application used to generate it. To search for a specific string in the payload, we can use the rule option content which lets us specify the data to be searched in ascii text or in hexadecimal. Now, we also want to be alerted when the admin user is logging in to the service, so we first analyze the traffic Using a sniffing tool which could be Snort itself or some ‘other like tcpdump, or Wireshark. From our analysis we need to find the pattern for admin login. The following is the output from topdump when the admin login was in process (see Listing 1) ‘As we see from the packet capture that there is a string usr admin which is being used as a command to login as admin. This string can be used as our pattern to detect admin login. We can write the following rule to search the pattern: a o@ ove rt se om se on ¢t A small tool that makes a big difference Authors, Florian Eichelbenger anc Ludwig Ertl can be Pere eee eee eens.BASICS Listing 1. Only TCP Payloadis displayed here 0201 9075 7372 2062 6466 €96e 0001 7565 6224 3739 Ga66 6076 6276 6866 6465 2693 33346 3500 7065 7264 3a77 7777 7800 0201 0075 66a €3¢6 0001 3738 6866 6276 Seat With this rule, Snort will skip the first 3 bytes of the payload and start searching the string usr admin within the next 9 bytes. This way the rule is also optimized as the search will not be carried out for the entire payload, 5 Relative Content os Matching To further reduce the possibility of false ain. 0634793 rnegi7user atnin login det sid:3000002) vevls This rule will trigger whenever the admin login happens, however there is also a good possibilty of it triggering at other times when this string is not being used as a command but as some data. This would be termed as a false positive. While choosing patterns, ‘we must keep in mind that our pattern is as much as possible unique to the required event. Reducing False Positives We will analyze the packets more closely looking for more patterns to reduce the possibilty of false positives. Firstly we can add the constant binary data preceding and following our previous string. Binary data can be represented in hexadecimal and enclosed within the characters. If the binary data was not constant in all the admin logins, then we need to look at other parameters. Let's say the string usr admin always starts after three bytes of some variable data. We can instruct Snort to skip the first three bytes of the payload and start matching the pattern string immediately from there. For this we need to use two content modifiers — offset and depth. Content modifiers affect how the content option works and have to follow the required content option for e.g. (content: usr admin; nocase}) will look for the string usr admin but ignore the case, here the nocase is a content modifier. »IHARING SNorT alarms, we can match additional pattern if the first match is successful. We have noticed that there is a string perm: which always comes 24 bytes ahead of the usr admin string (see Listing 2). The rule logic to match the second pattern will be same, however the keywords used here are different as now the match begins from where the previous match ended, The content modifiers used for this new pattern are distance, which is similar to offset, the only difference being that distance is an offset from the end of previous match whereas offset starts from the beginning of the Payload, and the other one is within which is similar to depth but works in conjunction with distance. e124; within: cogin detectec? 1d:3000002; xevs4z ) ‘There are many more ways to enhance the rule, maybe we can cover them in the next article. If you have to say anything about this article or have a wishlist for the next one, you can write to me and | shall surely consider. KISHIN FATNANI- CISSP, GCIH GOLD, GCFA, CCSE R70, CEH, LPT Kishin Is the founder of K-Secure (www.ksecure.net), an IT security training company. He has over 20 years experience in IT, main focus being security, and conducts many security trainings including Packet Analysis, Snort Rule Writing, Ethical Hacking, Forensic Analysis, Web Application Security, Check Point and more. He has also been a SANS local mentor for the GIAC certification GCFA and has some contributions to the development of Check Point courseware and certification His contact him at kishinfksecure.net Blog: kishinf.blogspot.com Facebook: www.facebook.com/ksecure Twitter: www.twitter.com/ksecure wychaking orgThe Challenge of — figgum2 Compliance Q.What are the sorts of challenges large corporations are facing with regards to compliance? KK: The challenges are many and this is because the compliance environment is becoming more complex. Business operations are becoming dynamic by the day, regulators are becoming more watchful when it ‘comes to data security and customers are becoming more demanding. A very simple way to know you've got a problem is to ask your CISO: How secure are we & as an organization. If you find him hemming-hawing or jving some vague response, then you really don't have a handle on your information security. Q. How are companies de: challenging environment? KK: What enterprises should aim to do is put in place a framework, which helps critical questions that provide a clear idea of where the organization stands with regards to information security. Our firm makes one 19 with this wwrwhaking.org rica ies der 0 ‘such compliance platform called as NX27K (http:// www.niiconsulting.com/products/iso_toolkit.htm!) = which helps companies comply with the multiple information security compliance frameworks and regulations, such as ISO 27001, PCI DSS, SOX, etc. Q. What should be the capabilities of such a platform? KK: The key feature of such a platform should be flexibility. And this is the key principle on which NX27K is built. You can modify what you see and put into your, asset register, or modify your risk assessment formula, modify the audit trail, the risk register, and almost every ‘aspect of the platform. Itintegrates with Sharepoint and Active Directory. Further, it adapts risk assessment ‘uidelines such as from NIST and COBIT's Risk-!T Q.Where can customers go for more information? KK: Customers can email us at products@niiconsu Iting.com or visit us at hitp://www.niiconsulting.com/ products/iso_toolkit.htm! for more information on this product and our other services and products. Certified Professional Hacker, 13th Dec 2010 to 17th Dec 2010 nep:/wue.security infcourses/ephv2.htont Certified Information Security Consultant, 13th Dec 2010 to 30th Dec 2010 tp:d/wermaisecuriyin/courses/eisc html Certified Professional Hacker - eLearning tp:/fisecurityn/elearning htm SNORT HARING |.BASICS Collection and Exploration of Large Data Why the use of FastBit is a major step ahead when compared with state of the art relational database tools based on relational databases. What you will learn... + Basis of building traffic monitoring applications + Data Collection and Exploration What you should know... + A basic knowledge of architecture and implementation of trafic monitoring tools + Abasicknowledge of TCPAP ollecting and exploring monitoring data is becoming increasingly challenging as networks become larger and faster. Solutions based on both SQL-databases and specialized binary formats do riot scale well as the amount of monitoring information increases. In this article 1 would like to approach to the problem by using a bitmap database that allows to implementation of an efficient solution for both data collection and retrieval NetFlow and sFlow NetFlow and sFlow are the current standards for building traffic monitoring applications. Both are based on the concept of a traffic probe (or agent in the sFlow parlance) that analyses network traffic and produces statistics, known as flows, which are delivered to a central data collector. As the number of flows can be pretty extremely high, both standards use sampling mechanisms in order to reduce the workload on bothof the probe and collectors. In sFlow the use of sampling mechanisms is native in the architecture so that it can be used on agents to effectively reduce the number of flows delivered to collectors. This has a drawback in terms of result accuracy while providing results with quantifiable accuracy. With NetFlow, the use of sampling (both on packets and flows) leads to inaccuracy and this means that flows sampling is very seldom used in NetFlow hence there is no obvious: mechanism for reducing the number of flows records, wlHARING SNORT while preserving accuracy. For these reasons, network operators usually avoid sampling data hence have to face with the problem of collecting and analyzing a large number of flows that is often solved using a flow collector that stores data on a SQL-based relational database or on disk in raw format for maximum, collection speed. Both approaches have pros and cons; in general SQL-based solutions allows users to write powerful and expressive queries while sacrificing flow collection speed and query response time, whereas raw-based solutions are more efficient but provide limited query facilities. The motivation is to overcome the limitations of existing solutions and create a new generation of a flow collection and storage architecture that exploits state-of-the-art indexing and querying technologies. In the following | would lke to describe the design and implementation of nProbe , an open-source probe and flow collector, that allows flows to be stored on disk Using the FastBit database. Architecture and Implementation nProbe is an open-source NetFlow probe that also supports both NetFlow and sFlow collection and, flow conversion between version (for instancei.c. convert v5 to v9 flows). Itfully supports the NetFlow v9 specification ‘so giving it has the ability to specify flow templates (ie. it ‘supports flexible netflow) that are configured at runtime when the tool is started (Figure 1). wewchaking orgcc] When used as probe and collector, nProbe supports flow collection and storage to either raw files or relational databases such as MySQL and SQLite. Support of relational databases has always been controversial as users appreciated the ability to search flows using a SQL interface, but at the same time flow dump to database is usually enable only realistic for small sites. The reason is that enabling database support could lead to the loss of flows due to the database overhead. There are multiple reasons that contribute to this behavior and in particularincluding: + Network latency and multi-user database access for network-based databases. + Use of SQL that requires flow information to be converted into text that is then interpreted by the database, instead of using an API for directly writing into the database + Slow-down caused by table indexes update during data insertion + Poor database performance when searching data during data insert. Databases offer mechanisms for partially avoiding some of the above issues, which includinge: + Data insert in batch mode instead of doing it in real time. + Avoid network communications by using file-based databases. + Disable database transactions. + Use efficient table format optimized for large data tables. + Not defining tables indexes therefore avoiding the overhead of index updates, though usually results in slower data search time. Other database limitations include the complexity of handling large databases containing weeks of data, and purging old data while stil accommodating new flow records. Many developers partition the database often creating a table per day that is then dropped when no longer needed. ‘The use of file-based databases such as SQLite offer a few advantages with respect to networked relational databases, as: + ILis possible to periodically create a new database (e.g. one database per hour) for storing flows received during that hour, this is in order to avoid creating large databases, + According to some tests performed, the flow insert throughput is better than networked-databases but still slower than raw flow dump. wwrwhaking.org Geb ee ed + In order to both overcome the limitations of relational databases, and avoid raw flow dump due to limited query faciliies, | decided to investigate the use of column-based databases and in Particular, of FastBit . Validation and Performance Evaluation Ihave used the FastBitlibrary forcreating an efficientflow collection and storage system. This is to demonstrate thal nProbe with FastBit is a mature solution that can be used on in a production environment. In order to evaluate the FastBit's performance, nProbe has been deployed in two different environments: Medium ISPs The average backbone traffic is around 250 Mbit/sec (about 40K pps). The traffic is mirrored onto a Linux PC (Linux Fedora Core 8 32 bit, Kernel 2.6.23, Dual Core Pentium D 3.0 GHz, 1 GB of RAM, two SATA II disks configured with RAID 1) that runs nProbe in probe mode. nProbe computes the flows and saves them on disk using FastBit. In order to reduce the number of flows, the probe is configured to save flows in NetFlow v9 bi-directional format with maximum flow duration of 5 minutes. In average the probe generates 36 million flows/day. Each FastBit partition stores one hour of traffic. Before deploying nProbe, flows were collected and stored in a MySQL database. Large ISPs nProbe is used in collector mode. It receives flows from 8 peering routers, with peak flow export of 85 K flows/sec. The collection server is a fast machine with 8 GB of memory, running Ubuntu Linux 9.10 server 64 bit. Each FastBit partition stores five minutes of traffic that occupy about 5.8 GB of disk space. A second server running Ubuntu Linux 9.10 server 84bit and 24 GB of memory is used to query the flow data. The FfastBbit partitions are saved to a NFS. mount on a local storage server. Before deploying Packet Capture Flow Expat aw Flos /MySOL/SOLie / Fast Figure 1 SNORT HARING IBASICS Probe, flows were collected using nfdump and each month the total amount of flow dumps exceeds 4 TB of disk space. The goal of these two setups is to both validate nProbe with FastBit on two different setups and compare the results with the solution previously used. FastBit vs Relational Databases Let's compare the performance of FastBit with respect to MySQL (version 5.1.40 64 bit), a popular relational database. As the host running nProbe is a critical machine, in order not to interfere with the collection process, two days worth of traffic was dumped in FastBit format, and then transfered to a Core2Duo 3.06 GHz Apple iMac running MacOS 10.6.2. Moving FastBit, partitions across machines running different operating systems and word length (one is 32 the other is 64 bit) has not required any data conversion. This is a good feature as over-time collector hosts can be based on various operating systems and technology; hence flow archives can be used immediately without any data conversion is a desirable feature. In order to evaluate how FastBit partition size affects the search speed, hourly partitions have been merged into a single daily, Partition. In order to compare both approaches, five queries can be defined: QU: sexzcr couns(+s,SUMIPRTS),SUN(BETES) FROM NETELOW + Q2: seuscr couwness + Q3: seize counrcey FROM NETELON GROUP BY cAVE_SRC_ + 4: sexect svm_sne avon suxconas) suminyres) AS. + 5: seuzer rev4_sec_A0oR, 14_sne_rowr, 1ev4 oor ‘Table 2. FastBitvs MySQL Query Speed results aren seconds) MySQL Pee etn Cre FastBit partitions have been queried using the fbquery tool with appropriate command line parameters. All MySQL tests have been performed on the same machine with no network communications between client and server. In order to evaluate the influence of MySQL indexes on queries, the same test has been repeated with and without indexes. Data used for testing washave been captured on Oct 12th and 13th (~68 milion flows) and contained ‘a subset of NetFlow fields (IP source/destination, port source/destination, protocol, begin/end time). The table below compares the disk space used by MySQL and FastBit. In the case of FastBit, indexes have been computed on all columns, Merging FastBit partitions does not usually improve the search speed but instead queries on merged data requires more memory as FastBit has to load a larger index in memory. In terms of query performance, FastBit is far superior compared with MySQL as shown in Table 2: + Queries that require access only to indexes take less than a second, regardless of the query type + Queries that require data access are at least an order of magnitude faster that on MySQL. + Index creation time on MySQL takes many minutes and it prevents its use in real life when importing data in (near-)realtime, and also indexes also take a significant amount of disk space. + Indexes on MySQL do not speed up queries, contrary to FastBit, + Disk speed is an important factor for accelerating queries. In fact running the same test twice with data already cached in memory, significantly decreases the query speed. The use of RAID 0 has demonstrated that the performance speed has been improved. Open Issues and Future Work Tests on various FastBit configurations have shown that the disk is an important component that has a major impact on the whole system. | am planning to explore the use of solid-state drives in order to see if the overall performance can benefit from it performance increases. A main limitation of FastBit is the lack of data compression as it currently Se CE compresses only indexes but a 208 26 a 234 6 03 029 e796 on 76 46 41033 ua a 572 os 1754 257 445, 281 «I HARING SNORT not data. This is a feature is planned to add, as it allows 15 os — disk space to be saved rence to reducereducing 557482 i the time needed to read the 473307 data, wewchaking orgThis article is the base for developing interactive data visualization tools based on FastBit partitions. Thanks to recent innovations in web 2.0, there are libraries such as the Google Visualization API that allow separating data rendering from data source, Currently we are extending nProbe adding an embedded web server that can make FastBit queries on the fly and return query results in JSON format. The idea is to create an interactive query system that can visualize both tabular data (e.g. flow information) and graphs (e.g. average number of flows on port X over the last hour) by performing FastBit queries. This way the user does not have to interact with FastBit tools at all, and can focus on data exploration. The use of FastBit is a major step ahead when compared with state of the art tools based on both relational databases and raw data dumps. When searching data on datasets of a few million records the query time is limited to a few seconds in the worst case, whereas queries that just use indexes are completed within a second. The consequence of this major speed improvement is that it is now possible to query data in real time and avoid updating costly counters every second, as using bitmap indexes It Is possible to produce the same information when necessary, Finally this work paves the way to the creation of new monitoring tools on large data sets that can interactively analyze traffic data in near-real time, contrary to what usually happens with most tools available today. This work is distributed under the GNU GPL license and is available at the ntop home page hitp:/www.ntop.org/ nProbe.html. The nBox appliance embedded withing a pre-installed ntop and nProbe software can be requested at www.wuerth-phoenix.com/nbox. LUCA DERI, FOUNDER OF NTOP Luca Deri was born in 1968. Although he was far too young to remember, the keywords of that year were freedom, equality, free thinking, revolution. In early 70s many free radio stations had birth here in Italy because their young creators wanted to have a way for spreading their thoughts, ideas, emotions and tellthe world that they were alive ‘n kickin’. The Internet today represents for him what free radio represented in the 705. He wrote his PhD on Component-based Architecture for Open, Independently Extensible Distributed Systems. Luca Deri is the founder of Ntop. wwrwhaking.org The new The flow-based network probe and monitoring appliance Who is using your network? What makes your network traffic? Who is consuming most of the bandwidth? Ifyou want to answer these questions you should become a nBox user. ‘© High-performance embedded NetFlow™M, ‘8 v5/v9/IPFIX probe ‘= Embedded Net Flow vs/v9/IPFIX collector ‘© 1Pv4, IPV6 and MPLS support '= Easy to set-up and configure ‘© No additional delay in both mirrored traffic and existing network ‘© User friendly web GUI for nProbe and ntop ‘© Multiple collector mode for load balancing or redundancy ‘© Firmware and packages upgrade via Internet '= All software reside on flash disk ‘© Ability to dump NetFlowTM flows on-disk or on Database Server KEY FEATURES. The new nBoxis a cooperation between ntop founder Luca Deri and Wirth Phoenix. Get more information at worw.wuerth-phoenix.com/nBox G5 — mia) Netw PFD rite abe t Comvee* ackets Packets Ga) a. Detelerator ux Kernel Packets t WUORTHPHOENIX worw.wuerth-phoenix.com/nBox wawentop org,ADVANCED Improving your custom Snort rules While it is easy to create a custom Snort rule, do you know if you are actually making a good one or not? This article introduces some common mistakes | find in custom Snort rules and the potential implications of those mistakes. What you will learn... + How fo measure the performance of your custom Snort rule set, and how to identify the "bad performers + How to use Snort’ fast_pattern keyword to your advantage + How to add mote descriptive information to your tues to improve the analysis process What you should know... + The Snort rule language + Abasicknowledge of TCP + How to install and Perl modules via CPAN on a *NIX operating system over the last ten years. Packet processing speed has improved, IP defragmentation and stream reassembly functions have evolved, the connection and state tracking engine has matured, but there is one thing that keeps getting left behind, Custom rule-sets. With each revision of Snort new features are added that enhance the detection capability and aid in packet processing performance of the Snort engine. Those enhancements not only open new avenues for detecting the latest bad stuff out there, they create an opportunity to improve the performance of older legacy rules you may have created many years ago. Unless your rules make good use of the current Snort language and are kept up-to-date, what once used to be a good rule could in fact turn bad. T he Snort IPS engine has changed substantially What isa bad Snort rule anyway? Because this article focuses on finding and fixing bad rules, before we look at any of them it would be wise for me to define what | personally call a bad rule. + Arrule that cannot or will not perform the function it's made for. Le. it won't catch the attack/event that the rule tries to find (False negative) + Arule that catches the wrong stuff (False positive) + Arule that has little meaning or value to the analyst (unk alerts) »IHARING SNORT + A rule that wastes processing time and effort to achieve its goal (performance hog) The first two points in the list (dealing with false positives and false negatives) will always need to be addressed on a case-by-case basis, however when it comes to the last two points there are some important concepts you can follow that will substantially improve custom rules regardless of what attack or event they are designed to catch. ‘There are many books and online resources available that discuss the Snort rule language in depth, and a full introduction is far out of the scope of this article However, to enable me to present some common rule problems, I need to introduce the basic building blocks of a Snort rule. ASnort rule is made up of two parts, a rule header and. a rule body. The rule body follows the rule header and is Listing 1. An example "bad" Snort rule AL NEY 80 \ wewchaking orgei eeu surrounded by parentheses, The header is pretty easy to understand as it reads close to natural language. The rule header consists of an action, a protocol specification, and the traffic that is to be inspected, The rule body (shown in Listing 1 in blue) is made up of a selection of rule options. A rule option consists of a keyword followed by one or more arguments, For example in the above rule there is a content keyword, with an argument of os:easbenthasfo0 This rule instructs Snort to look for the text ondend » in all packets flowing out of the network to TCP port 80 that are part of an established TCP session. Giving rules more meaning to an analyst The above rule in Listing 1 is a simple example of a bad rule. Regardless of how much processing load it may introduce to an IDS engine, or how many alerts it could generate, just by looking at the rule source we can see its bad. This rule lacks some of the most important information that could give any value to its existence and operation ~ an understanding of what's happening oon the network, and what it means if it generates an alert When describing or teaching the Snort rule language, I ike to group rule options together to describe them in three categories based on their function within the rule, + Detection Options: Keywords that test for the presence of specific things in network traffic. This could be any type of text, binary data, packet header values, regular expressions, or decoded application data, These are the keywords that control if an alert is generated on a packet or not. Example Snort keywords: content, pere, ipopts, tt, flowbits, flow + Metadata Options: These are keywords that are interpreted by the engine for alert organization, Example Snort Keywords: sid, metadata, rev + Analyst Options: Keywords that are used by the rule writer to convey information to an event analyst who is investigating the event. Example Snort keywords: msg, classtype, reference, priority While writing a rule it Is important to understand that any event it generates may need to be understood by other people. If a security analyst is presented with an event like the below which was generated from our previous bad rule, what do they do to respond to the event? 2.268,0.212854 wwrwhaking.org Well after the analyst has likely scratched their head for a moment, and then wondered what on earth .atoo is, they will probably end up Googling for oxussbeutbsaroo in an effort to understand what this alert means to them, Is this a serious event? Is it somebody else's problem? Should | start panicking? It is common to have a different group of people researching and writing rules vs. those will who deal with the security events they may raise, and if this isn't the case for you today it may well be in the future, At the time of rule creation only the rule writer really knows what she or he Is looking for, and the implications to the network if the traffic is found. It is therefore critical for this information to be passed on to the event analyst within the rule itself. Unless a rule is correctly explained, how can a writer expect an analyst to be able to react accordingly? Let's expand on my simple oxsesdseesvaaes> example from earlier by providing some more theoretical scenario information (Listing 2). Note the addition of three new rule options: A classification type, an overriding priority qualification, and a couple of useful references. With these extra Tule options added an analyst dealing with the event now knows that oxdesdbeesneatoo IS in fact a low-priority Trojan, associated with CVE:2010-99999, and a related toa specific company's product. ‘These seemingly minor additions make massive returns in respect to the security event analysis and remediation process. Sometimes the simplest changes provide the greatest value. Identifying and optimizing slow rules that are wasting your CPU cycles So while fixing the analyst information problem is pretty simple, identifying suboptimal rules in terms of Listing 2. The badrule, now improved with far more infermation to help an analyst ‘$0u_ NET any > SEXTERNAL NET 40 \ teojan-activicys \ ‘eve, 2010-98999; \ rl, nesp:/Mnycons ssa:100001) rev:2, SNORT HARING:ADVANCED computational overhead is a litle more of a technical process. To make this challenge possible Snort can kindly provide us feedback of how the system functions in relation to the current configuration and network traffic being inspected, There are a couple of useful configuration lines that, ‘can be added to your snort.conf to provide performance feedback about how the detection engine is performing, Today | will focus on the output provided by porte cong profiler Adding this ponte rutes configuration directive to your snort.conf will enable performance profiling of your snort rule-set. At exit, Snort will output to STDOUT a list of the top N (specified here as ten) worst Performing rules categorized by the total time taken to check packets against them. This data can also be written to a text file of choice, and many other sort methods are available, Check the snort manual for full details, Note Snort must be compiled with - enable the performance profiling capability. Before starting to inspect the performance output, it is vital to understand that all of the data we see is dependant on two distinct variables: ng to + The current configuration running (including the rule-set) + The network traffic that is detection engine inspected by the When testing and tweaking anything as complex as an IPS rule-set for performance, | find it imperative to isolate and work on only a single variable at a time, By focusing my tests on a large sample of network traffic stored in PCAP files that is representative to where the sensor operates, | can tweak my rules for performance against this static data-set. When | think have optimally tweaked any rules, I can then move to test against live traffic. ‘An example of rule profiling output is shown in Listing 3, and each data column is explained below. + Num: This column reflects this rule's position number in regard to how bad the rule performs. Here the top (number 1) reflects the rule that is responsible for consuming the most processing time (cots: ) + SID, GID, Rev: The Snort ID, Generator 1D, and Revision number of the rule. This is shown to help us identify the rule in question in our rule-set, + Checks: The number of times rule options were checked affer the fast_pattern match process (yes, that bit is bold because it is important), + Matches: The number of times all rule options match, therefore traffic matching the rule has been found + Alerts: The number of times the rule generated an alert, Note that this value can be different from Matches due to other configuration options such as alert suppression. + Microsecs: Total time taken processing this rule against the network traffic + AvgiCheck: Average time taken to check each Packet against this rule. + Avg/Match: Average time taken to check each packet that had all options match (the rule could have generated an alert) + Avg/Nonmatch: Average time taken to check each packet where an event was not generated (amount of time spent checking a clean packet for bad stuff) The two values a rule writer has some level of control over are the number of checks, and how long it took to perform those checks. Ideally we would like to have low figures in all columns, but decreasing the Checks count is the first important part of rule performance tuning. To be able to tweak our rule to affect this value, we need to first understand exactly what Checks represents. Introducing the fast_pattern matcher When Snort decides what rules need to be evaluated against a network packet, it goes through two stages before starting its in-depth testing functions. Listing 3. Sample Snort rule profiling output HAHING wewchaking orgei eeu 1) Protecol, Port number and service identification Snort optimizes the rule-set into protocol, port and service (application protocol) based detection rule- buckets. Note that service based buckets are only used when Snort is compiled with -enable-targetbase, and an attribute table is loaded. For example, if inbound traffic is destined to arrive at TCP:80 (HTTP), there isn’t much point in running it though the rules associated with SMTP (TCP:28). The packet is assessed against the rules in the TCP. 80 bucket. The same decision is also made related to source port and service metadata. Snort also has an extra rule-bucket for the any any rules. These are the rules that use the value anyas both the source and destination port numbers. All packets are also checked against the any any rules as well after being assessed against their particular port/service based rule bucket. 2) Packet content (fast_pattern check) After identifying what rule bucket(s) this packet should be assessed against, a pre-screening content check known as the « .e=s match is applied for all rules in the bucket(s) For any Snort rule to raise an event all rule-options in that rule must match. Applying a rss:_pacte:n check process allows Snort to quickly test packets for the presence of a st content string (a single content: value) required to generate an event. The goal of this test is to Quickly identify all packets that have any possibility of alerting after all of the rule options are tested. if a packet doesn't match the csot_ps:cern check, there is absolutely no point in running more computationally intense checks against it. Because the match has failed, we know that at least one of the rule options will not match, and an alert will never be generated. Number of Checks This brings us back to the important Checks value. The number of checks is the number of times a tule is assessed after both the protocot/port/ service identification and sast_pse complete. The more irrelevant packets that we can exclude with these two steps, the lower the number of checks will be, the more optimal and focused the rule will be, and the less time will be wasted performing in-depth assessment of packets that will never generate an event. » processes are Identifying the optimal content check for fast_pattern By default Snort selects the string from the longest content keyword (measured in number of characters) wwrwhaking.org for use in the cast. sattezs test, The design rationale behind this is simple — the longer a content check, the more unique it will likely be, therefore less packets will inadvertently match it. Although this is commonly the case, there are times when the rule writer will have a different opinion based on knowledge and experience. Looking at the rule in Listing 4, Example.com is the longest content check (eleven characters), and by default will be used for the ras+_sactern check. The other content CST-ID-001 is however less likely to be found in network traffic, especially if your company name just so happened to be Example.com. It is therefore wise to tell Snort to use this better value for the rast_psctern check with the sat potter modifier keyword, Following the pactern check, each rule option is then tested in the order that it appears in the rule. Finally the source and destination IP addresses are tested to check if they match those defined in the rule header. Only if every check made is successful is the rule action (such as alert) taken. If any of the tests fail to match, no further checks are made on the packet against that rule, therefore it is advisable to place quick tests such as SEXTERNAL NET SHTTP_§ Potential Data Leak: Honey Token cSt SNORT HARING:ADVANCED 19 5. Dumbpig.plin action varefubontl ercblem(s) found with References: tp:/snortorg httpy/ur-sourcefre blogspot.com! httpy/leonward.wordpress.com/dumbpig/ DumbPig can be found at itp:/code.google.com/ p/dumbpig/, it is written in Perl and uses Richard Harman's very Useful rarso:;ssort rer) module to parse arule-set, Listing 5 shows the tool in action. In this example ddumbpig pl is reading in a rule file called ras.rstos, and has identified two problems with rule sid:1000008. The first problem shown in Listing 5 is in my experience very common. A rule-writer has created a rule for DNS traffic that is commonly found on both TCP and UDP port 53. Rather than create two rules (one for TCP and one for UDP), the rule writer has used the IP protocol in the Snort header, but has also specified @ port number. Because the IP protocol doesn't have ports (port numbers are a transport layer construct), this value is ignored. The end result is that every packet regardless of port will be checked for this content. This is very sub-optimal to say the least. The 2IHARING second problem with this rule is that itis missing extra analyst data that will provide more value to any alerts raised, Summary The Snort rule language is simple to pick up, and ina similar way to any other language It Is easy to fall into ‘some bad habits. Hopefully this article has introduced some simple suggestions that will improve any of your in-house IDP rules in respect to their performance and usefulness. LEON WARD Leon is a Senior Security Engineer for Sourcefire based in the UK. He has been using and abusing Snort and other network detection technologies for about ten years and hates referring to himself in the third-person. Thanks go to Alex Kirk (Sourcefire VRT) and Dave Venman (Sourcefire SE) for sanity ‘checking this document. wewchaking orgpeered fern Insecure Websites in DMZ Still Pose a Risk Level of Trust Normally, awebsiteis considered to bea partof the untrusted ‘outer perimeter ofa company network infrastructure, Hence, system administrators usually put a web server in the DMZ, part of a network and assume the information securiy risk from the website to the network is mitigated. However, several industry security standards have been imposed to protect the public infrastructure such as webs servers and name servers in addition to the services directly subjected to the standard application scope. Why is i so important to, protect your website even if itis not closely connected to your critical data infrastructure? Social Impact Humans are the weakest ink in the chain of a company's security. The experience gathered during more than 5 years of penetration testing shows that almost no large-scale companies can resist a social-engineering attack vector. In companies which have more than 30 employees, a penetration tester or a real intruder can pretext, deceive, and easily persuade at least 10% of the available employees to open an attachment or follow a link to the malicious website containing an exploit pack and a Viral payload. Basic countermeasures include restricting network access to all websites but whitelsted sites, Which includes your own website, or simply educating the employees. So, what happens when an intruder gains access to the website? The following list highlights what can be done with a web server located in DMZ: + Inject an exploit pack and payload into the main page or create malicious pages + Send spam and scam letters to the company ‘employees inviting them to visit a malicious page at the website + Install a rootkit and sniffer to maintain access and get all password inputs by system administrators or website maintainers + Modify links from legitimate sites to malicious ones, for instance, to redirect Internet bankin Onthe‘Net [1] OWASP Testing Guide ~ htav/aww.owasp.org/indexphp/ Category OWASP Testing Project [2] OWASP Development Guide — http:/wuwowasporg/ index php/Category OWASP Guide Project [B1 InformzaschitaSC(QSA, PA-QSA)~ htp:/vwinfosec.ru/en wwrwhaking.org VHdopm3auta (Cucrembtili uetrerpaTop link to http:/fibank.yOurbank.ru instead of ‘ttp:// ibank yourbank.com + Pivot client-side payloads through the web server in the case of networks with restricted Internet access. This list includes only those risks related to successful network penetration. In addition, there are business image risks such as defacing, modifying sensitive public information (e.g. exchange rates at bank's website, payment credentials at some charity company's website, phone numbers etc.), or denial of service by deleting everything and bringing the web server down. Ways to Protect There are several methodologies to assess website security and mitigate risks connected with the website. One of the most popular is the OWASP Testing Guide [1], which includes more than 300 checks and addresses almost all known web vulnerabilities. The PCI Data Security Standard refers to the top 10 most widespread ‘vulnerabilities in the software, called the OWASP Top Ten, and is a basic requirement for any website dealing with credit card payments. For developers, there is also ‘a Development Guide [2], the goal of which is to prevent mistakes affecting security, To companies willing to protect their websites and networks, Informzaschita offers the following services: + Complete website assessment according to OWASP Testing Guide (300+ checks) + Express assessment according to OWASP Top Ten and deployment of Web Application Firewalls for small businesses or companies falling under PCI DSS requirements + Complete PC! PA-DSS assessment for companies developing payment applications + Automated security web and network scanning MARAT VYSHEGORODTSEV, Information Security Assessment Specialist at Informzaschita [email protected], (+7 495) 980-2345 wwwinfosec.rufen SNORT HARING |.ADVANCED An Unsupervised IDS False Alarm Reduction System - SMART Signature-based (or rule-based) network IDSs are widely used in many organisations to detect known attacks (Dubrawsky, 2009). A common misconception about IDSs is that they are Plug-and-Play devices that can be installed and then allowed to run autonomously. In reality, this is far from the truth. What you will learn. + The imitations of 0S tuning + The basic concepts and characteristics of SMART system + The benefits of SMART What you should know... + Basics of Intrusion Detection Systems + Basic syntax of Snort rules The need for alarm reduction Depending on the quality oftheir signatures, they can generate a significant volume of false alarms. Even when the alarms are genuine, the sheer number that can be generated with aggressive attacking traffic (0.9. Denial of Service attacks, vulnerability scans, malware propagation) can be a problem. This is especially the case if the IDS has been configured to generate an alert each time a rule matches. Tuning is often necessary to address the problems of superfluous and false alarms. However, if done recklessly, it can increase the risk of missing attacks. On another note, when attacks span multiple stages, it would be useful to have a mechanism of aggregating and grouping together all alarms relating to the same activity. This not only enhances detection, but also the efficiency of analysing and validating alarms. In order to address the issues above, alarm reduction systems are needed. Alarm reduction is a process that analyses the intrusion alerts generated by IDS, filters the false alarms and then provides @ more concise and high level view of detected incidents cost Figure 1. Framework of false alarm classification model »IHARING 10S sensor TDS sensor TDS sensor Events Tats SMART (SOM K-Means Alarm Reduction Tool) SMART is an automated alarm reduction tool, designed for Snort IDS. It extends the functionality of Basic Analysis and Security Engine (BASE) (BASE, 2009), a popular front-end for Snort, by providing a more holistic view of detected alarms. The system comprises of two stages; the first stage is responsible for removing superfluous alarms, whilst the second stage distinguishes true from false alarms, by ‘observing trends in their occurrence. The idea behind SMART is not to replace human analysts, but to inform alarm validation and IDS tuning by identifying the most relevant candidates (alarms and rules) for Figure 1 depicts the proposed classification model. Specifically, it shows that data is collected from IDS sensors, and stored in a database. The system then retrieves the data from the database and classifies them by extracting the attributes from the alerts and feeding them into the unsupervised SOM-based clustering system, wewchaking orgCO aL atau Acie Luau The classification process consists of the following phases: 1, Feature extraction — The system uses several attributes extracted from the alert database, which are considered effective to correlate alerts generated from a single activity. The extracted data are then normalised since the value of the data are varied depending upon the type of attributes used. 2. Alarm aggregation (first stage correlation) - Given a set of input vectors from the first phase, the system is trained unsupervised in the second phase to map the data so that similar vectors are reflected in their arrangement. The distance between two input vectors is presented on the map, not by their absolute dissimilarity (which can be calculated), but the relative differences of the data properties. The objective is to group alerts from the same attack instance into a cluster. 3. Cluster analysis - The result of the classification is further evaluated to attain a set of attributes from each cluster created in the previous phase (ie. the first stage correlation), Building accurate and efficient classifiers largely depends upon the accuracy of the attributes, which are used as the input data for the classification, Seven alert attributes (as shown in Table 1) were chosen to represent the value of each input vector in the next classification (i.e. second stage correlation). Two out of the seven attributes, namely the frequency of alarm signatures and the average time interval between the alerts each day were computed. These features are considered to be the most relevant in terms of influencing the magnitude of the alert signatures. 4, Alert classification (second stage correlation) — The final classification is carried out based upon the attributes extracted in the third phase. The main objective of this stage is to label the alerts into true and false alarms, thus reducing the number of alerts before being presented to the administrator. The underlying architecture of our proposed alarm classifier that illustrates the four phases of the classification process is presented in Figure 2. Evaluating SMART In order to evaluate the effectiveness of SMART, a set of experiments has been conducted. The dataset that was used for the experiments was collected on a public network (100-150 MB/s) over a period of 40 days, logging all traffic to and from the organisation’s web sorver. It contained 99.9% of TCP, and 0.1% of ICMP traffic. The effectiveness of reducing false alarms In SMART is compared against conventional tuning methods. False alarm rate by tuning Tuning is @ popular technique of reducing false alarms, and it is based on the adaptation of the IDS configuration to suit the specific environment where the IDS is placed (Chapple, 2003). This often involves modifying preprocessors, removing (or modifying) rules prone to false alarms, or modifying variables. The first phase of the experiments involved running snort in default configuration and validating the generated alarms to identify the most suitable candidates for tuning. The following three rules were the most suitable, as they triggered the most false alarms: WEB-IIS view source via translate header This event is categorised as web application activity, which targets the Microsoft IIS 5.0 source disclosure vulnerability (Snort, 2010c). Surprisingly, this signature alone had accounted for 59% of the total alerts, in which approximately 1,970 alerts were generated per day by this event. Although the signature was created to detect a Microsoft IIS source disclosure vulnerability exploitation attempt, Le 1 The interpretation and data collection methods ofthe alarm attributes for second stage correlation ERT FEATURES | DESCRIPT ‘The service port number Indicates ifthe alarm contains a well-known port number, or unknown service calty ofthe alerts. There ate 3 types of alert priority, namely 1st, 2nd, and 3rd, multiple signatures are No of alerts Total numberof alerts grouped in one cluster Noofsignatures Total number of signature type ina cluster Protocol ‘Type of trafic from event triggering the alerts Port number ports. Alert priority ie found in a cluster, the priority value for each signature could be added together. Time interval Time interval between events from a particular signature No of events The number of events1 in which a particular alert signature is triggered within a day “One event equals to a number of alerts from a single signature, which are triggered by a particular activity. wwrwhaking.org SNORT HARINGADVANCED in this case all generated alarms were related to normal traffic. When examining the snort rule, it does not seem proficient enough to detect this type of event. It appears to be very loosely written, by searching for a particular string in the packet payload (in this case, Translate: N) Since the Translate: fis a valid header used in WebDAV application (WebDAV, 2001), the rule tends to trigger a vast volume of alarms. If the rule gets modified to search for the GET command in the content, itis likely that false alarms would be reduced. The attack is launched by requesting a specific resource using HTTP GET command, followed by Translate: fas the header of HTTP request. in this case, a tuning can be performed by modifying the signature rule to: WEB.IIS view source via translate header ~ Tuned signature rule 2 xeferencerbugtrag, 14764; xeference:bugtrag, 15782 0494, by eeserenceiness ase Indeed, when snort was run again with the modified ruleset, this rule had effectively eliminated 95% of the initial false alarms. WEB-MISC robots.txt access Although this event is raised when an attempt has been made to directly access robots.txt file (Snort, 2010b), it can also be raised due to legitimate activities from web robots or spiders. A spider is software that gathers information for search engines by crawling around the web indexing web pages and their links. Robots.txt file is created to exclude some pages from being indexed by web spiders (e.g, submission pages or enquiry pages). As web indexing is regular and structurally repetitive, this activity tends to cause a superfluous amount of alerts. In this study, ‘Networ alarm | ‘ate CO ‘acreaaton Flee ‘Custer ‘Alar PosttverTve = = lassfeaton Negative at Foatre Selection Figure 2. Architecture of alse alarm classifier “HARING approximately 23% of total alerts (approximately 750 alarms per day) were related to this activity, and they were all false alarms. In order to modify this rule to exclude normal web spider activity, the source IP addresses would need lo be examined, in order to verily their authorisation in accessing the Robots.txt file. This approach, however, seems to be hardly feasible to deploy. Of course, identifying all authorised hosts from their source IP addresses is impractical and dangerous for exploitation. Specifying such a large number of IP addreses can be a problem. Also, the mere fact of allowing specific hosts to access this file could be exploited in order to bypass detection, As such, event thresholding was used instead (Beale and Caswell, 2004). As robots.txt access requests generate regular and repetitive traffic, a limit type of threshold command is the most suitable tuning in this case. Such a threshold configuration would be as follows: The rule logs the first event every 60 seconds, and ignores events for the rest of the time interval. The result showed that approximately 10% of false alarms had been effectively reduced. This indicates that tuning can only reduce a very insignificant number of false alarms from this event. ICMP L3retriever Ping is an event that occurs when ICMP echo request is made from a host running L3Retriever scanner (Snort, 2010c). Quite a few alerts were generated from this event; contributing to 8% of the total alerts. This figure indicates that approximately 250 alerts were generated by this rule every day. Surprisingly, there were no malevolent activities detected following the ICMP traffic. In addition, normal ICMP requests generated by Windows 2000 and Windows XP are also known to have similar payloads to the one generated by L3Retriever scanner (Greenwood, 2007). In view of this issue and given that no suspicious output detected following these ICMP requests; these alerts were labelled as false positives. The only method that can be deployed to suppress the number of false positive triggered from this event is by applying event suppressing or thresholding command. Instead of using limit type of threshold command as previous signature, this rule utilised both type of command to log alerts once per time interval and ignore additional alerts generated during that period: wewchaking orgrea alert icmp SENPERNAL NET any -> SHOME_NE? any (nsg:"1CMe Lineteiaver Ping*;seode:0 itype:sy content: NOPQASTUVABCOBAGHH"; depth:327,zeference:acachnida, 311; classtypesattenpted-recon; threshold: type both, track by_svey count 3, seconds 60) si 65 2e054) The threshold is written to detect brisk ICMP echo requests by logging alerts once per 60 seconds after seeing 3 occurrences of this event. This experiment has also proved that the event thresholding can successfully reduce up to 89% of the false alarms generated by this activity. Overall, fine tuning has been effective in reducing false alarms. However, several limitations exist: 1. The procedure increases the risk of missing noteworthy incidents - Suppressing the number of alerts generated can also create a possibility of ignoring or missing real alerts. For example, a malicious user can hide his/her action within the excessive number of alerts generated by using a spoofed address from web spider agent. In fact, looking for an overly specific pattern of a particular attack may effectively reduce the false alarms; however, this method can highly increase the risk of missing its range. A skilful attacker can easily alter and abuse the vulnerability in various ways as an attempt to evade the IDS. 2. Tuning requires a thorough examination of the environment by qualified IT personnel and requires Bae cee Luu a frequent updating to keep up with the flow of new vulnerabilities or threats discovered False alarm rate by SMART The following stage of the experiments involved analysing the snort alerts with SMART. This experiment. presents the results of SMART classification, which is run every two hours, using only four hours alerts from the private dataset. This is due to increased memory requirements of running our unsupervised alarm reduction system for larger data. So, instead of running ‘one correlation for the entire data set, SMART runs a correlation over a particular time period (e.g. every ‘one or two hours). Figure 3 shows the maps of the correlations. The classification reveals that about 78.8% of false alarms have been identified in the first map (left), whilst 96% of them have been detected in the second, mappings (right), as shown in Figure 3. Those alarms located in the upper portion are labelled as true alarms, whilst the lower portion is for the false alarms. It is, notable that our system has shown promising result in filtering all hectic and unnecessary alerts triggered by the IDS. For example, the alerts from WEB-IS view source via translate header and WEB-MISC robots.txt access signatures, which had caused 82% of false alarms from the entire data. In addition to the private data set, the system has also been tested using publicly available data set, DARPA 1999. The experiment also shows promising results, e = Figure 3. SMART classification result using private data set wwrwhaking.org 2eustes SNORT HARING.ADVANCED References 1. BASE (2008), Basic Analysis and Security Engine (BASE) Project, htp:/base secureideas.net/ 2. Beale, J. and Caswell (2004), Snort 2.1 Intrusion Detection, 2nd edn, Syngress, United State of America, ISBN 1931836043 3. Chapple, M. (2003), Evaluating and Tuning an Intrusion Detection System, Information Security Magazine, http:/searchsecurit ylechtarget.com/tp/1,289483,sid14 gci918619,00.html 4, Dubrawsky, | (2008), CompTIA Security+ Certification Study Guide: Exam SYO-201 3E: Exam SYO 201, Study Guide and Prep kit, 3rd edn, Syngress, United States of America ISBN 1597494267 Greenwood, B. (2007), Tuning an IDS/IPS From The Ground UP, SANS Institute InfoSec Reading Room, 2007, httpy/ wunw.sans.org/reading_room/whitepapers/detection/tuning-idsips-ground_1896 Snort (2010a), WEB-IS view source via translate header, http://www.snort.org/search/sld/10422¢=1 Snort 20106), WEB-MISC robots.txt access, htip:/iwww snort org/search/id/18522"=1 Snort 20100), ICMP L3Retriever Ping, http:/wwwsnort.org/search/sid/466?r=1 WebDAV (2001), WebDAV Overview, Sambar Server Documentation, htpy/ww.kadushisoft.com/syshelp/webdav.htm reducing up to 95% and 99% of false alarms in the first and second classification respectively. The system appears effective in filtering the false alarms triggered by @ noisy traffic such as ICMP traffic (ICMP Ping and Echo Reply) and web-bug alerts, which have formed the highest number of false alarms. Overall, SMART has been effective in detecting false alarms, such as the redundant and noisy alerts raised by ICMP traffic. In fact, it is also proved that the system outperforms the traditional tuning method In filtering the WEB-MISC robots.txt access alerts. In other words, the issue of subjective rule suffered by common tuning method can be addressed using the GINA TJHAL She holds a BSc Computer Science from the University of Wollongong, Australia (2005), and an MSc in Information ‘System Security from the University of Plymouth, UK (2006). She is currently a PhD candidate in the Centre of Security, Communications & Network Research at University of Plymouth, UK. Her current research interests include network intrusion detection and prevention, pattern classification, neural network and data mining. MARIA PAPADAKI ‘Maria Papadakiis a lecturer in Network Security, at University of Plymouth, UK. Prior to joining academia, she was working as a Security Analyst for Symantec EMEA Managed Security Services (MSS), UK. Her postgraduate academic studies include @ PhD in Intrusion Classification and Automated Response (2004), and an MSc in Integrated Services and Intelligent Networks Engineering (2000), University of Plymouth, UK. Her research interests include intrusion prevention detection and response, network security ‘monitoring, incident prioritisation, security usability, and security education. Dr Papadaki is a GIAC Certified Intrusion ‘Analyst, and is a member of the GIAC Advisory Board, as well 4s the British Computer Society. Further details can be found cat www plymouth. ac.uk/escan. »IHARING proposed system. More than 0% of false alerts from WEB-MISC robots.txt access can be identified by ‘SMART. Summary Performing a fine-tuning to reduce false alarms is not a straightforward task. If not done properly, there is a possibilty that the system might miss real attacks SMART is an automated alarm reduction system, which helps fitering false alarms generated by Snort IDS. It has advantages over the conventional tuning method, Unlike tuning, the system does not require any prior knowledge of the network and protected systems. In addition, it provides a higher level of alert information to the administrator by aggregating alerts from the same attack instance and validates the accuracy of Snort IDS. Acknowledgments The authors would ike to acknowledge the contribution of Prof Steven Furell and Dr Nathan Clarke In the work presented inthis paper. wewchaking org~~ lin eee a erolaal= an ate ate = ESS SoM LV ella og iia =n = Tecotmtcliare Eat Na ine Me oe Tie ao iale Kee pert clog Ete Toc) stole ETO sata) intial] elec CMMs MC) ol eo eatoa Maca lac helele) foc. Errata oil la Mas Rca Male are Tiara pee =a) =a) n= oer ec Ma Ica Naa ae ca a iscal nip POE RAM eed =a a encase Kaspersky Lab Premier Partnen ESET Premier Partner Corel Pl Adobe Silver Solution Partnen Autodesk Gold Partnen ABBY Lal arc aaeriatat eee Meee Me Noel o Meo Malo \Vi-V-Wpletv-) p= Ceo ee RO Ree ra aie sea i= ny) © | aoe Mig loa ac Roeia ecg Met ial ee goal || Sos-1>1 py Grote Wate la Reta la oleg = ae leg =| Pine cocioa lac tev couacect gloat isi l| WoVinin |) Sa ocge tae ta Matcto ake) = Negra aaa s ities and 20 countries. ao aae IN a isnbaL ical stoi Coie a Mier ip stop a No ic) RONaaceraeliaatetnelt Retoiaal Ushers a tae