Security Analytics Dr. Char Sample February 2014 Agenda • Definitions & Tools • Trends – Cloud – Big Data – TIOT – Behavior • Summary • Q&A February 2014 2 SECTION 1: DEFINITIONS & TOOLS February 2014 3 Definitions • Security Analytics – can be thought of as the process associated with developing insights to an environments actual security based on the inputs collected from the various security components. • Translation – Using data to inform a decision February 2014 4 Why Do We Care? • Security professionals are increasingly be required to be conversant in them. – Why? • Old solutions do not work. • The realization that the “easy button” is still elusive. • Analytics are where the “rubber meets the road” • Money! Analytics (for the time being) require human interactions and thought, generating JOBS. February 2014 5 Why Analytics Now? • Drivers – – – – Cloud computing Big Data Existing technology shortcomings New devices. • Mostly, this effort is being driven by the need to see beyond the host and the LAN. • Situational awareness. February 2014 6 Why Analytics Now? • Changing roles and expectations for security professionals. – Inadequacies of scan and patch approach have been exposed. – Dynamic network environment • Exposes C&A shortcomings • Exposes architectural assumptions and shortcomings. – Attackers are becoming more stealth. – No clear solution leader has emerged, so DIY. – Uniqueness of environments. February 2014 7 Security Analytic Components • Tools that we use to assist in gaining insight include: – – – – – – – Protocol Analysis Traffic Analysis Flow and Metadata Logs (system, session, pcap, metadata, flow data) System tools Statistical models (Markov, Bayes, Fuzzy Logic, etc.) Mathematical models (Clustering algorithms, neural networks) February 2014 8 Tools & Data • Tools & Data Used for Analytic Creation – Protocol Analysis – IETF standards • • • • February 2014 Log data Signature data Packet capture data (pcap) – tcpdump Anomalous data 9 Tools & Data • Tools & Data Used for Analytic Creation – Traffic Analysis – Data quantities • • • • • • Flow data – IPFIX, others. Metadata – IPFIX with statistical knowledge, others. Log data – syslog, application logs, etc. Signature data – IDS signatures Packet capture data (pcap) – tcpdump Anomalous data – Scripting languages - perl, python February 2014 10 Tools & Data • Protocol Analysis – Commonly used commands vs less used commands (debug, trace, post vs put) – Commonly used parameters with commands (executable commands, wrong type parameters) February 2014 11 Tools & Data • Traffic Analysis* – Unexplained changes in volume to existing IP addresses and ports – New destination IP addresses, and ports February 2014 12 Tools & Data • Signature data – Commonly used. • NIDS, some firewalls, AV, anti-malware – Highly accurate with known attacks • Performs a basic pattern match, can be done by hardware. • Low false positives, but high false negatives • 10% effective at detecting new or 0 day attacks – Best use is in lowering false positives from AD, and keeping out less sophisticated attackers. – Ineffective against nation-state and other sophisticated attackers. – Removal of this type of data leaves analyst vulnerable to information overload. February 2014 13 Tools & Data • PCAP data – Entire packet • • • • Preferably put together into sessions. Storage is a problem Legally admissible evidence Provides details that other data types are incapable of providing. – Removal of this data type results in a lack of detailed understanding of events seen by flow and meta data. February 2014 14 Security Analytic Tools • DNS lookup tools: nslookup, dig, whois • Mapping tools: IP_address to ASN, traceroute, arp table, netstat February 2014 15 Security Analytic Tools – DNS Info • DNS tool: nslookup February 2014 16 Security Analytic Tools – DNS Info • DNS tool: dig February 2014 17 Security Analytic Tools – DNS Info • DNS tool- whois: Internet domain name and network number directory service February 2014 18 Security Analytic Tools – DNS Info • DNS tool: whois (more from query) February 2014 19 Security Analytic Tools – Routing Info February 2014 20 Security Analytic Tools – Routing Info • Tools – traceroute (limited use): print the path between hosts February 2014 21 Security Analytic Tools - ARP • Connectivity tool- ARP: address resolution display and control February 2014 22 Conclusions • Several tools are available to assist in determining the nature of the activity these tools are diagnostic tools. – Connectivity tools: ping, traceroute – Routing tools: ASN – IP address mappings – DNS tools: dig, nslookup, whois – Log data February 2014 23 END SECTION 1 February 2014 24 SECTION 2: TRENDS February 2014 25 Cloud & Big Data • The “Cloud” provides both the source of our data, and the ability to analyze it. • Cloud as defined by NIST – IaaS, PaaS, SaaS – Characterized by dynamic provisioning – Shared resources February 2014 26 Defining the Cloud February 2014 27 Defining the Cloud February 2014 28 Items to Consider • Virtual Machines • Network Infrastructure with multiple parties – Recall the earlier discussion on global routing and how routing works. – The importance of understanding the ASN relationships. • The role of layer 2 data – Remember the arp command • Multiple IP addresses associated with the same arp address indicates virtualization (cloud) • Even when arp addresses are unique there are certain ranges of arp addresses that are “fake” and are used with virtualization • Security products are layer 3 based so they typically do not see Layer 2 behavior. February 2014 29 Items to Consider • Problems in virtual environment require analysis of the CSP’s environment. – If you are in a public cloud sharing space with other entities how will you know if their space has not been compromised? – How can you be sure that allocated resources are clean? – It is not enough to know that the security apparatus in your virtual environment is working, that is only one piece of the puzzle. February 2014 30 Items to Consider • Layer 2 February 2014 31 Items to Consider • More on layer 2 data – Bugs in the hypervisor – Provisioning and de-provisioning. • Is data really wiped? • How are requests authenticated and processed? February 2014 32 Cloud IaaS • IaaS Analytic Basics – Understand the routing infrastructure of the cloud, all tenants and each cloud site. – Consider having data encrypted before it is sent to the cloud. – Understand the path between the CSP sites • Know the peering agreements between CSP locations. • Understand the routing between tenants. Most likely layer 2 technology will be used and virtual addresses will be set. February 2014 33 Cloud - IaaS • IaaS Analytic Basics – Routing • Look for the sudden appearance of new routers in tenant space. • Look for changes in routes, especially tenants suddenly having new traffic come through their space. • Look for changes in the CSP’s virtual router routes or the switch that would allow cross tenant access. February 2014 34 Cloud - IaaS • IaaS Analytic Basics – If CSP is managing DNS check delegation – If CSP is also running DNSSEC extra work must be done with DNSSEC key management issues, check the DNSSEC delegation dig +sigchase – Examine DNS data from inside the zone, cloud partition, also examine from the CSP and finally from an external point. • Verify views. • Disable recursion from external ANS. • Check DNS BIND logs, look for strange commands, executable statements, and other anomalous activity. • Examine for evidence of tunneling between tenants: queries from neighbors are a good clue. • Take note of queries to/from fast flux sites, especially those managed from other CSPs. February 2014 35 Cloud - IaaS • IaaS Analytic Basics – DNS • Look for changes in DNS Server’s arp address. • Look for cross tenant DNS/TCP traffic. • Look for changes in tenant zone information that are out of sync with normal updates. • Look for changes in authoritative nameservers (SOA) advertising. • Look for changes in MX records that direct away from known tenant mail hubs. • Look for tenants advertising multiple domains, domains that they may not own, or fast flux behaviors. • Look for tenants that set up recursive resolvers for other tenants. • Look for tenants that forward requests to other tenants. • Look for tenants that set up DHCP servers for other cloud tenants. February 2014 36 Cloud - IaaS • IaaS Analytic Basics – DNS • DNSSEC analytics – Look for changes in ZSK that are out of sync. – Look for changes in ZSK following new access to the server. – Look for activity surrounding changes to key management procedures or users for both ZSK and KSK – General • When dealing with a truly distributed attack it is possible that the IP addresses will change but the MAC addresses will be either the same or from the same pool of “fake” addresses. This would indicate that the attack is coming from a virtual environment. February 2014 37 Cloud - PaaS • PaaS – How is PaaS being utilized? • Programmers are a common reason for the need for PaaS • Test data can also reveal information about the organization. How will that data be protected? • If programmers are using: – Is DEBUG available? – How is the development environment wiped? – What about object permanence? • On the development platform: – Check security logs for signs of activity on the cluster ports. – Check for odd changes in activity on cluster hosts and ports. – Check for evidence of misconfigurations. February 2014 38 Cloud - PaaS • PaaS - Hadoop February 2014 39 Cloud - PaaS • PaaS Analytics – Look for changes in activity levels on the following ports: 8020, 50010, 50020, 50030, 50060, 50070, 50075, 50090, 50100, 50105. – Look for activity from new hosts on the above ports. – Look for signs of imposter NameNode attempts – Look for new mac address for NameNode – Look for appearance of new DataNodes, especially in existing clusters. – Look for strange commands and/or parameters. – Check authentication logs. February 2014 40 Cloud - SaaS • SaaS – Run lint checkers on new apps – Code review? – Signatures – Protocol Abuse – Command abuse February 2014 41 Cloud – SaaS Example February 2014 42 Big Data • The dynamic behavior of the cloud requires less traditional analysis to understand. – Unstructured, continuous data requires new methods – Finding patterns within Big Data is key to analytic development in the cloud – Architecturally, this will impact choices on security solutions. February 2014 43 Big Data • Big Data – Volume – obvious – Variety – many new types of data are available in “big data” that haven’t been used by security in the past, and this data is unstructured, e.g. streaming. – Velocity – dynamic relationships, everything from flow, to proxies, to pcap come in. Analysts cannot keep up with the data. – The 3 V’s of BD makes flow data extremely important! February 2014 44 Big Data February 2014 45 Big Data • What does the Hadoop explanation mean, why do we care? – There are several potential exploit areas that must be considered. • The distributed nature of the cluster • The each individual cluster inter-node issues • Communications between the NameNode and the DataNodes • Additional copies of data used to support HA • Software vulnerabilities within MapReduce & various libraries. February 2014 46 TIOT • The Internet of Things – NSA has declared 2014 the year of TIOT – Supervisory Control and Data Acquisition (SCADA) & Other Embedded Devices • Protocols: ARP, UDP, ICMP, DHCP,TCP, PPP, SNMP • Look for new entries into the arp table, look for new ports and new destinations. February 2014 47 Behavior • Recognition that technical methods alone are insufficient. – Understanding attacker “patterns of thought” – Understanding and anticipating “automatic thought processes” • • • • February 2014 Group behaviors Game theory Cultural influences* Psychological influences 48 Summary • The security products and services industry has failed to keep up with the changes in the industry. • Consultants are now being asked to define and create security analytics. • A good understanding of tools and capabilities will assist in most cases along with understanding of user requirements. • Cloud security adds a new dimension since other tenants, the supply chain and the platforms must all be considered. • TIOT will stress the signature and AD models. • Big Data requires thought models and new visualization techniques. February 2014 49 END SECTION 2 February 2014 50 Q&A February 2014 51 THANK YOU! Dr. Char Sample csample@cert.org charsample50@gmail.com +1 301.346.9953 February 2014 52 Cloud • PaaS - Hadoop – Namespace is stored in RAM on the Name Node and it contains: • Filenames • Location information – Name Node acts as the controller and queries the data nodes on behalf of the client. • Replication of data is standard and 3 copies is the default but this value is configurable. • Look for unexplained activity on HDFS ports. • Look for unexplained hosts accessing Hadoop ports. February 2014 53 Routing February 2014 54 Routing • Obtain info on BGP peers February 2014 55 Cloud - PaaS • HDFS February 2014 56 Cloud - PaaS • HDFS February 2014 57 Cloud • SaaS – Check permissions. – Check activity levels • Traffic volume • Hosts accessing. • Changes in top users, or users using new features. – – – – Check libraries Evidence of object permanence Evidence of cross tenant communications using apps Evidence of one tenant acting as a server for another tenant. February 2014 58 Layer 7 Analytics • Analytics can be done at all layers of OSI • IP addresses and DNS provide basic background (against APTs) • Application and Web Server logs provide data capable of determining intent • These new logs sources are pushing the Big Data movement forward. Analysts are now asked to understand a far wider breadth then they have before in a much shorter period. February 2014 59 Security Analytic Tools • DNS tool – nslookup: query Internet nameservers interactively – Easier than dig, but provides less information – Built in for Windows, Unix and MacOS – Does provide information on various entities • • • • • February 2014 SOA MX A IN PTR CNAME 60 Security Analytic Tools – DNS Info • DNS tool: dig February 2014 61 Big Data • Hadoop keying on common objects. – Framework for storing • Big Data • Distributed server clusters • Distributed analysis applications in each cluster. Nodes in the cluster used for storing data are known as data notes. – Apache – Rather than rely strictly on h/ware for HA Hadoop uses software. • This means that multiple copies of data have to be made in order to support HA • Library is used to detect failures at the application layer, delivery of HA services occurs on top of the cluster. • HDFS is a data warehouse which differs from a database. • Accessing Hadoop for queries relies on MapReduce – NameNode – DataNode February 2014 62