Effective Security Practices Present and Future Chris Misra

Effective Security Practices: Present and Future Chris Misra, UMass - Amherst Marty Schulman, Juniper Networks Joe St. Sauver, U. Oregon Jack Suess, UMBC

Security Working Groups • EDUCAUSE/Internet 2 task force working groups: – Security and Awareness – Risk Management – Effective Practices • SALSA - Sponsored by Internet 2 and focusing on security in high performance and advanced networks and providing advice on leading edge technology issues.

Why Effective and not Best? • Higher education institutions have very different characteristics and are too diverse for the onesize fit that best implies • Network security is evolving and security practices will have a limited lifespan • Goal for today - identify how SALSA and the Effective Security Practice work group are working towards enhancing security

Effective Security Practices Initiative • The goal of the initiative is to identify and publicize practical approaches to preventing, detecting, and responding to today’s security problems. Released online guide in January 2004, • Eoghan Casey developed this guide in consultation with campus security officers • Effective practices working group meets via biweekly conference call to identify, solicit, develop, and review new practices.

Effective Security Practices Guide Focus Areas Online at the www. educause. edu/security/guide Contents include • Education, Training and Awareness • Risk Analysis and Management • Security Architecture Design • Network and Host Vulnerability Assessment • Network and Host Security Implementation • Intrusion and Virus Detection • Incident Response • Encryption, Authentication & Authorization Presently we have 25 practices in the guide

ESPG Highlights Evolution of Security Practices

SALSA Security Table • Found in SALSA report, Pages 8 -13 http: //apps. internet 2. edu/sals/files/20031108 -wr-sals-v 1. 1. pdf • Summarizes operational and performance considerations for 18 network-based security technologies and 8 host-based technologies • The table also classifies each technology as preventative, detective, reactive or analytical in nature.

Security Design and Architecture • Effective practices guide has 3 sections devoted to the topic: – Security design. Focuses on network technologies. Security strategies from Ga. Tech, Michigan, U. Wash. , and GMU. – Security Implementation (stage 1) - standards, configuration management, patching, and authentication. – Security Implementation (stage 2) - security policies and techniques for managing desktops

Architecture WG Marty Schulman marty@internet 2. edu

Short Bio • Started as technical aid at PSC – When backbone was 56 kbps… • Several network jobs since – SURAnet, BBN, Bell Atlantic, Sprint, Cisco • Currently Juniper Federal Systems Chief Technologist • Majority loaned to Internet 2 – Ex officio SALSA member

Security Tools In Theory • Many items in security toolbox – Identify: shared secret, 2 factor, biometric – Authenticate: PPTP, SSO, IPSec – Control/Isolate: Co. S, VPN, stateless or stateful filtering – Audit: syslog, flow export • They are not inherently good or bad…

Security Tools In Practice • Campus environments vary widely – Widely varying security policies – Widely varying budgets – Firefights and growth by emergency • Temporary changes become permanent – Varying states of deployment • Vendors try to help. . . – Unique terminologies and marchitectures

Consequences • The network “funhouse” – You can’t get there from here – You can’t even see what “there” looks like • Impeded progress – Inconsistent terminology – Frustrated operations – Harder to take advantage of I 2 mutual trust

Net-arch Working Group • Net – Will not consider exclusively Layer 7 – Some techniques may involve applications • Architecture – Assumes well-defined boundaries – Identifies the functions and devices in which they live – Identify guidelines/design rules

Group Plans • Technology, not policy – You decide what pieces are appropriate – Guidance to enable avoid pitfalls and enable federation • First deliverable – Ontology – May reference 3 rd party works

Volunteers Needed • Draft homepage – http: //security. internet 2. edu/arch/index. html – Working charter – Mailing list • Additional contacts – Steve Olshansky – Charles Yun

Effective Security Practices: SALSA-Network Access Control Internet 2 Member Meeting Arlington, VA 20 April 2004 Christopher Misra University of Massachusetts Amherst cmisra@oit. umass. edu

Network Access Control • Networks are valued resources to the R&E community • Security incidents are (always) on the rise – Polybot/Phatbot – IRC Botnets • Authenticated and authorized access to network resources is critical in many environments – Policy enforcement

Network Access Control • Methods and techniques currently exist to provide network access control to campus networks. – Identification/Authentication – Device Registration – Authorized access • Inconsistent terminology and functional description

Network Access Control: Effective Practices Guide • How has the community extended access control techniques to effectively manage security incidents • Case study: – Network Registration System Scanner – University of Connecticut – “Using Net. Reg Scan kept the Blaster and Welchia infection rate on our student network to about five percent of all hosts. ”

Network Access Control: What next? • There is no standardization among currently deployed network access control techniques in R&E networks. – Taxonomy • Network access control is a component of a security architecture • Given some future network access control architecture, can we federate it? – Accommodate visiting scientists

SALSA • An advisory group consisting of technical experts from the higher education community – advise on leading edge technology issues – provide prioritization, and set directions in the security space. – SALSA will be future-oriented and state-of-the-art in nature, focusing on high performance and advanced networks.

SALSA-netauth Working Group • Consider the data requirements, implementation, integration, and automation technologies associated with understanding and extending network security management related to: – Authorized network access • By user and/or device – Style and behavior of transit traffic • proscriptive and descriptive – Forensic support for investigation of abuse

Network Security Management • How do we legitimately register hosts. • How do we use network diagnostic data to identify or remediate malicious activity. – Proscriptive • Firewall, packet shaper, etc – Descriptive • Active – Vulnerability scanning, patch verification • Passive – Things like netflow, IDS. Subject to privacy rules • How can we combine these to improve our security management capability

Network Security Management • The scope and scale of recent security incidents have shown the value of access control. • Requires a rich set of data across disparate data sources to provide a cohesive network information base for: – device location – incident remediation

Network Security Management • Tracability. – How do we find and stop malicious activity from compromised systems • Device state validation. – What conditions must be met by a device before allowing access to the network? – Is the device vulnerable to known exploits? • Guest Access – Given network access control policy, how do you accommodate visiting scientists. – Roles

Network Security Management Several cases: • Trying to identify a known compromised system. – Legitimately registered system – Legitimate owner likely does not condone activity – Legitimate owner is not trying to evade detection • Trying to identify intentional malicious activity. – Possibly a non-registered system – Activity is intentional by legitimate owner – Legitimate owner is trying to evade detection • Repudiation of suspected malicious activity

SALSA-netauth Working Group: Initial activities • Investigation of requirements and implementations of network database and registration services in support of network security management; • Investigation of extensions to these services to proactively detect and prevent unauthorized or malicious network activity. • Analysis and proposal toward a pilot and eventual implementation to support network access to visiting scientists among federated institutions. • Analysis of security applications that may

SALSA-netauth Working Group: Volunteers needed • Draft homepage – http: //security. internet 2. edu/netauth/index. html – Draft charter – Mailing list • Additional contacts – Steve Olshansky – Charles Yun

Email Effective Security Practices: 5 Concrete Areas To Scrutinize Internet 2 Member Meeting Arlington VA, April 20, 2004 Joe St Sauver, Ph. D. University of Oregon Computing Center joe@uoregon. edu http: //darkwing. uoregon. edu/~joe/emailsecurity/

Email Security and Its Role in Your Overall Network Security Plan • Many of the network security threats you face are directly tied to email security issues. • Unfortunately, because email is considered to be rather “mundane” or plebian, email security issues sometimes get short shrift. • In point of fact, email security deserves extra attention because it is the one application that is truly ubiquitous, and is truly mission critical. • Our goal is to highlight five concrete areas to scrutinize during our ten minute long slot. • We’ll assume a Unix-based email environment.

#1: Encrypt Your POP & IMAP Traffic • Hacker/crackers love to sniff ethernet traffic for usernames and passwords. • One of the most common sources of usernames and passwords on the wire consists of clear text POP and IMAP logins to campus mail servers. • Most popular POP and IMAP clients and servers now support TLS/SSL encryption, including Eudora, Outlook, Entourage, Mozilla, Mulberry, OS X’s Mail program, etc. (See the recipes at http: //micro. uoregon. edu/security/email/ ) • If you are NOT requiring encrypted POP and IMAP logins, the time has come to do so.

Controlling Other Plaintext Password Exposures • If you also offer a web email interface, be sure it is also always encrypted (runs via “https”) too. • Require ssh (not telnet or rlogin) for any access to Pine or similar command line email programs. • Replace ftp with scp or sftp, etc. • Work to eliminate any legacy shared (rather than switched) network segments (switched ethernet is not a panacea, true, but it can help) • Secure. ID/Crypto. Card-type token based auth systems may also be worth testing/evaluation • Encourage use of GPG (http: //www. gnupg. org/ )

SMTP Auth With STARTTLS • While you’re encrypting POP and IMAP traffic, you might as well also require SMTP Auth (RFC 2554) over a TLS encrypted channel as well. See: www. sendmail. org/~ca/email/auth. html • If you do deploy password based SMTP Auth, be SURE that you require strong user passwords (check ‘em with cracklib). Spammers will try exhaustive password attacks against servers using SMTP Auth in an effort to remotely relay (e. g. , see: http: //www. winnetmag. com/ Articles/Print. cfm? Article. ID=40507 ). Watch your logs/limit bad password attempts/tarpit abusers!

#2. Neutralize Viruses and Worms • Your users face a constant barrage of inbound viruses, worms and other dangerous content. Remember all the viruses “fun” of Fall 2003? [http: //www. syllabus. com/news_issue. asp? id= 153&Issue. Date=9/18/2003 (and 9/25/2003)] • Depending on your email architecture, you may be able to run each message through an AV scanner such as Clam. AV (a GPL-licensed Unix antivirus product, see: http: //www. clamav. net/ ) • If/when you do find viruses, please do NOT send non-delivery notices to forged message body From: addresses! (see http: //www. attrition. org/ security/rant/av-spammers. html )

Attachment Defanging/Stripping • If you can’t run a antivirus gateway product on your mail server, you should AT LEAST “defang” all executable attachments by having procmail stick a. txt onto the end of the original filename. [Attachments that are particularly likely to contain dangerous content (such as pifs and scrs) should get stripped outright from incoming messages]. See http: //www. impsec. org/ email-tools/procmail-security. html for a defanger • Be sure to spend some time thinking about how you want to handle zip files, passworded zip files with the password included in the body of the message alongside the zip file, . rar files, etc.

Users Still Need Desktop Anti Virus Software, Too • While you will likely do a good job of blocking viruses sent through your central email servers, users do still need a desktop AV product to deal with viruses coming through other email servers, infested web pages, peer to peer applications, instant messaging, Usenet, IRC, CIFS, etc. • When site licensed, commercial desktop A/V products can be surprisingly affordable. • Faculty, staff and students must use an desktop A/V product at work and at home (see free home options at: http: //www. pcworld. com/howto/ article/0, aid, 113462, tk, wb 122403 x, 00. asp )

Spyware • At the same time you deal with desktop antivirus requirements, be sure you also handle spyware. Spyware includes things such as web browser hijacking programs, key stroke loggers, long distance dialer programs, etc. You might think that antivirus programs would also handle these type of threats, but they usually don’t. • Recent estimates are that ~5% of hosts may be infested (See: http: //www. newscientist. com/ news/news. jsp? id=ns 99994745). • Antispyware reviewed: http: //www. pcmag. com/ article 2/0, 1759, 1523357, 00. asp (2 Mar 2004)

Your Users Should Also… • Be running a current version of MS Windows, or an alternative OS (Mac. OS X, Linux, *BSD, etc. ) • Apply all available service packs and critical updates (check for updates to MS Office, too!); enable automatic Windows Updates. • Use a personal software/hardware firewall • Users should routinely backup their system • Consider a system file integrity checker (cc. uoregon. edu/cnews/fall 2003/sysintegrity. html) • Use a strong password for their desktop system (particularly for Administrator accounts!) • Avoid using risky applications (P 2 P, IM, etc. )

Create a “Virus Resistant” Email Culture • A key determinant of the level of problems you have with viruses is your local “email culture”… -- Are non-institutional email accounts common? -- Do users routinely send plain text email only, or are attachments used even for short notes? -- Do users tend to employ a simple command line email program (such as Pine), or a more complex email program that’s tightly coupled to the underlying operating system (like Outlook)? -- Do users have a sense of healthy skepticism (regarding VISA phishing, 419 scams, etc)? -- See http: //www. columbia. edu/kermit/safe. html

#3. Manage Spam (Yes, Spam IS a Security Issue) • You probably are already taking steps to control spam, simply because spam now typically amounts to 75% of inbound mail (see: http: //www. postini. com), however spam is also a security issue. See: -- “Your computer could be a ‘spam zombie’” http: //www. cnn. com/2004/TECH/ptech/02/17/ spam. zombies. ap/ -- “Spammers, Hackers Increasingly Feed Off Each Other” http: //www. techweb. com/wire/story/ TWB 20040212 S 0009

Coping With Spam • There are many different ways to try to manage spam, but the two most popular mainstream approaches are: (1) to scan messages (including the message’s contents) using a tool such as Spam. Assassin, or (2) to block messages coming from insecure hosts and known spam sources via DNS-based blacklists (possibly augmented by local filters) • Other approaches (whitelisting, challenge/ response, hashcash, rate limits, collaborative fitlering, reputation systems, etc. ) all have fundamental issues that limit their applicability.

Spam. Assassin • By applying a variety of scoring rules (see http: //www. spamassassin. org/tests. html) to each incoming message, Spam. Assassin determines the likelihood that each message is spam. Typically, messages that look spammy get filed in a spam folder, while messages that look nonspammy get delivered to the user’s inbox. • The biggest issues with Spam. Assassin are (1) it requires that all messages first be accepted, then assessed and filed or discarded, (2) it relies on publicly-disclosed message characteristic heuristics for its filtering efficacy, and (3) it may be too hard for non-techy users.

DNS Black Lists • The alternative approach, which we prefer and recommend, focuses on where messages are from. • Message from a known spam source? Message from a known open relay or other insecure host? Block that traffic when the bad host tries to connect • Sites using DNSBLs often use run multiple lists, such as MAPS RBL+ (http: //www. mail-abuse. org/), Spamhaus SBL+XBL (http: //www. spamhaus. org/), and NJABL (http: //www. njabl. org/ ) • Arrange to download and run copies of any DNSBL zones you use on your own local DNS servers. • www. oag. state. tx. us/oagnews/release. php? id=413 • Tarpit info… http: //www. benzedrine. cx/relaydb. html

Be Sure You Allow Users to Opt Out of Your Default Spam Filtering • As a “pressure relief” valve, be sure to have a mechanism that allows users to opt out of your default spam filtering should they want to do so. • Here at UO, users can create a. spamme file in their home directory (either from the shell prompt or via a web-based request form) to signal that they “want out” of our default spam filtering. Every hour we look for those files, and adjust filters accordingly • If you do a good job of filtering, usage will be rare: as of 3/30/2004, 7 of 30727 UO student accounts have opted out, as have 38 of 13151 faculty/staff (plus 5 role accounts and 10 mailing lists)

AOL’s Latest Anti-Spam Technique (Controversial, But Apparently Effective) • AOL blocks spammers' web sites http: //www. washingtonpost. com/wp-dyn/ articles/A 9449 -2004 Mar 19. html “America Online Inc. has adopted a new tactic against spam: blocking its members' ability to see Web sites promoted by bulk e-mailers. ” • AOL reports drops in both e-mail & spam volume http: //www. clickz. com/news/article. php/3328841 “From Feb. 20 th to March 17 […] AOL delivered 37 percent fewer e -mails to spam folders, from 178 million to 113 million. Member spam complaints dropped by 47 percent, from 12. 4 million to 6. 8 million. ”

#4. Protect Your Deliverability (to AOL Users and Elsewhere) • Important mail that you send to your students and other folks may not be getting through… -- “[…] mail sent via UCLink/Listlink mailing lists to yahoo. com addresses is being blocked. ” http: //www-uclink. berkeley. edu/ cgi-bin/display/news -- “For several months, [Duke] was unable to send and receive emails to and from China…” http: //www. chronicle. duke. edu/vnews/ display. v/ART/2004/01/16/4007 df 2 ebfe 88 -- “Mail from IU to AOL blocked” http: //www. bus. indiana. edu/news/ View. News_Items_Details. asp? newsitemid=471&newsareaid=6 -- “After receiving a report indicating that no RAMS (Rutgers Automated Mass-mailing System) email messages were apparently making it into hotmailboxes, we decided to do a quick check to see if this was indeed true. Sure enough, the mail was not delivered to the mailbox with standard (default) mail filter settings in place. ” http: //camden-www. rutgers. edu/ RUCS-Camden/Announce/newsspring. 04. hotmaillink. html

AOL Scomps • One easy way to see if your users are emitting problematic email is to ask to receive AOL “scomps” (spam complaint reports) for your network blocks. See: http: //www. nanog. org/mtg-0310/spam. html • Caution: you may have infested systems that are spamming AOL users (and ONLY AOL users) which you’re unaware exist. If you haven’t been getting scomp reports previously, beware, the initial volume may be a little overwhelming… • I have reason to believe that other major ISPs will soon begin offering scomp-like spam reports

Secure Your Own Servers/Networks • We all know that insecure hosts, open SMTP relays, open proxy servers, exploitable formmail scripts, insecure ethernet ports and open wireless access points are Bad Things, right? (c. f. http: //darkwing. uoregon. edu/~joe/jt-proxies/) • Improving server security is now a global issue: http: //www. ftc. gov/opa/2004/01/opsecure. htm • Are you running a security scanner/auditing tool such as Nessus (http: //www. nessus. org/)? • Are you running a network intrusion detection system such as Snort (http: //www. snort. org/) or Bro (http: //www. icir. org/vern/bro-info. html)?

Other Things to Check/Do to Preserve Your University’s Email Deliverability • Are your mail servers on any DNSBLs? Check http: //www. openrbl. org/ • Are your hosts showing up in SANS reports? Drill down at http: //isc. sans. org/reports. html • Do you have an RFC 2142 -compliant abuse@ reporting address, or are you listed on http: //www. rfc-ignorant. org/ • Are you purchasing connectivity from spammerfriendly ISPs? See http: //www. spamhaus. org/sbl • Do your mailings follow emerging industry standards? http: //www. isipp. org/standards. php

If You Offer Institutional Mailing Lists… • All subscriptions to mailing lists must be confirmed by the requesting subscriber • Do NOT involuntarily put ANY users on ANY list (beware of the threat of “intraspam”!) • Anything except plain text that gets sent to a list should get stripped • Set list defaults to be reply-to-sender rather than reply-to-list by default • Prevent random harvesting of list memberships • Be sure to prevent harvesting of any online email directory you may offer, too!

#5. What About Filtering Port 25? • It is increasingly common among commercial broadband ISPs to filter customer port 25 traffic, forcing all inbound or outbound email to go through the provider’s canonical SMTP servers. By doing this, “direct-to-MX” spam from infected computers can be prevented, and infected customers can be identified from their message volume, and promptly disabled. • This sort of filtering of port 25 is explicitly discussed in RFC 3013 (“Recommended Internet Service Provider Security Services and Procedures”) at section 5. 4

Some Internet 2 Schools Have Filtered Port 25, Either Campus-Wide or For a Subset of Users (or Have Plans to Do So) • • • Buffalo: http: //cit-helpdesk. buffalo. edu/services/faq/ email. shtml#2. 2. 6 CWRU: http: //tiswww. case. edu/net/security/smtp-policy. html MIT: http: //web. mit. edu/ist/topics/email/smtpauth/matrix. html Oregon State: http: //oregonstate. edu/net/outages/index. php? action=view_single&outage_id=214 TAMU: http: //www. tamu. edu/network-services/smtp-relay/ University of Florida: http: //net-services. ufl. edu/security/ public/email-std. shtml University of Maryland Baltimore County: http: //www. umbc. edu/oit/resnet/faq. html#smtp-current-policy University of Missouri: http: //iatservices. missouri. edu/ security/road-map. html#port-25 (as of June 30, 2004) WPI: http: //www. wpi. edu/Admin/IT/News/networkingnews. html# newsitem 1059685336, 32099,

If You Do Decide to Filter Port 25… • If you do decide to filter port 25 traffic (except for traffic from your authorized SMTP servers), be sure you filter outbound AND inbound port 25 traffic. Why? Spoofed traffic from spammers “dual-homed” to a colo/dsl/cable ISP plus your compromised host/dialup, and who are sourcing packets from the colo/dsl/cable ISP with your compromised host’s/dialup’s IP addr. • If you really want to lock down unauthorized mail servers, be sure to also pay attention to 465/tcp (SMTPS) and 587/tcp (see RFC 2476), and also plan/decide how you’ll handle travelers (VPNs? )

An Alternative to Locally Filtering Port 25 • One alternative to locally filtering port 25 is “hinting” (via ptr/in-addr DNS entries) about groups of hosts that should probably not be sending email “direct-to-MX. ” For example: *. wireless. indiana. edu *. user. msu. edu *. resnet. purdue. edu *. dhcp. vt. edu Folks “out there” can then block smtp from those sort of hosts (or not) as they deem appropriate. • Avoid DNS naming schemes that require “midstring” wildcarding (dialup 67. example. edu)

DNS “Hinting” is Becoming Common in the Commercial ISP Space… *. adsl-dhcp. tele. dk *. cable. mindspring. com *. client. comcast. net *. customer. centurytel. net *. dial. proxad. net *. dsl. att. net *. dynamic. covad. net *. ppp. tpnet. pl • Consistent naming would be nice (but isn’t likely)

Another Option: Sender Policy Framework • SPF allows mail servers to identify and block forged envelope senders (forged “Return-path addresses”) early in the SMTP dialog by doing a simple DNS-based check of a site’s text record. • Many major providers/clueful sites are now publishing SPF records, including AOL (~24. 7 M subscribers), Google, GNU. org, Oreilly. com, Oxford. ac. uk, Outblaze (>30 M accounts), perl. org, SAP. com, spamhaus. org, w 3. org, symantec. com, etc. • What about your university? host –t txt example. edu

SPF Implementation Issues • Note that adoption of SPF can be done “asymmetrically” – you can publish your own SPF record but not query others, or vice versa. • If you’re used to email forwarding, get used to email rewriting (see the FAQ mentioned below) • Roaming users will develop a sudden interest in VPNs and/or authenticated remote access • You should know that here are competing approaches (such as MS’s Caller-ID). SPF implementations can also do Caller-ID queries • Want more information? http: //spf. pobox. com/ (the FAQ there is particularly helpful)

Thanks For the Chance to Talk Today! • Are there any questions?