Context Aware Firewall Policies Ravi Sahita Priya Rajagopal, Pankaj Parmar Intel Corp. June 8th 2004 IEEE Policy (Security) ® Overview Background Motivation Policy goals (example) Intrusion detection->Host<-firewalling Management SAFire Milestone conclusions •2• Communications Technology Lab Background Why firewall? Defense in depth against software flaws (software complexity increasing) Control over services accessed/exposed Control over information flow across boundaries (platform or network) Needed: Increased proactive response instead of reactive •3• Communications Technology Lab Policy goals (example) Track flow only if the session is initiated by client By default, restrict all traffic other than allowed services control traffic Create transient filters for the negotiated data flows On the negotiated port, restrict access to specific allowed commands/capabilities for that service When transferring data, block/flag suspicious content (so that it is checked) before it reaches apps All traffic that causes invalid protocol state transitions must be blocked proactively •4• Communications Technology Lab Advantages of host based FWs Visibility into internal traffic – Can protect against internal attacks Smaller number of flows, More state per flow – Decreased load on aggregation points Enable finer access control in a mobile environment – Carry your security Can use end-to-end protocol properties Allow true end-to-end encryption of traffic which would otherwise be proxied by the network devices •5• Communications Technology Lab IDS -> Host <- FW Context aware packet analysis (user, app, protocol, OS aware) Application layer gateways IDS complexity Firewall complexity End-point has this context information Traffic preprocessors, heuristics TCP level Stateful filtering Protocol analysis Stateless packet filtering blind signature detection Attack complexity Attack complexity •6• Communications Technology Lab Complex management Infrastructure firewalls are needed Host FWs=>number explosion, but valuable Make security policies easier to map without sacrificing functionality Make components tend towards autonomous behavior Make it easier to correlate events across hosts and infrastructure •7• Communications Technology Lab Why SAFire? What are the sub-elements of such packet analysis Allow building finer grain network access control policies Rich enough to keep up with new network services/changes Local remediation Abstraction of FW / IDS rules for a host •8• Communications Technology Lab Capabilities identified Flow state table management Application layer rules Pattern manipulation Outsourcing policy decisions Reuse of definitions Dynamic rule management •9• |---------HOST CONTEXT--------| Packet data extraction and filtering Communications Technology Lab Sequence of steps Express application protocol in a DFA Map protocol states to the Generic PSM Extract transition rules from the normalized PSM naming <src, event, dst, action> Map to SAFire primitives (using tools) • 10 • Communications Technology Lab Generic Protocol States FIN STOR|OK Extn * -{SYN-ACK} Suinit ACK SYNACK RETR| OK Extn PORT Sinit FIN Sde STOR| Not OK Extn * - {FIN} Sctd FIN Ste r m RETR| Not OK e xtn ACK ACTIV E FTP CONTROL TRAFFIC STATE TRANSITION DIAGRAM Sabor t * FILE CONTENTS M ALICIOUS * -{SYN-ACK} *{FIN} ACK Suinit SYNACK Sinit Sde FIN * - {FIN} Sctd FIN Ste r m CLEAN FILE FIN ACK Mapped to protocol specifics ACTIV E FTP DATA TRAFFIC STATE TRANSITION DIAGRAM • 11 • Communications Technology Lab Rule processing X Save State in Flow State Table Extract Packet Data Is Field =? Y Z Extract Packet Data Is Field ? Extract Packet Data Is Field X? T Get state from Flow State Table Extract Packet Data Is Field =X? Save state in flow table Extract Packet Data Is Field =Y? Extract Packet Data Is Field =Z? Extract Packet Data Is Field =? Extract Packet Data Is Field =Y? Extract Packet Data Is Field =T? Get state from flow table • 12 • Communications Technology Lab Implementation Local Firewall Configuration Application SAFire script in XML Remote Mgmt. Station SAFire Parser Static Filter Rules IOCTL Calls PSM Rules PSM Database Static Rule Mgr. PAE Core Packet Classifier Flow State Table Static Filters Filter Database Transient Filters • 13 • Communications Technology Lab Conclusions United model can comprehend HIPS+FWs Language extensibility = parallel progress Model allows security policy verification across implementations Minimal tradeoff is processing overhead for mapping and translation Context information on the host can be leveraged for finer access control Initial prototype shows minimal delay from user POV Communications Technology • 14 • Lab Thank you! Questions/Comments to ravi.sahita@intel.com • 15 • Communications Technology Lab