UCM Security: Challenges and Best Practices Brian “Bex” Huff Chief Software Architect
Oct 19, 2014
UCM Security: Challenges and Best Practices
Brian “Bex” HuffChief Software Architect
2
Purpose
Outline what you need to protect your content• How UCM fits into your security policy
• Use risk-management to minimize cost
Discuss risks and known vulnerabilities• General network briefly, UCM in detail
Present countermeasures to mitigate risks• General network briefly, UCM in detail
3
Brief History of Internet Security
Early days• Little security except governments, banks, and telcos
• Most software was far too trusting
First virus (1983), worm (1988), and trojan (1200 BC)• Networks must defend against malware
• Sparked battle between cryptographers and US government
Tech boom: • http://www.BURN-PILES-OF-MONEY.com
• Spent on security, but didn’t understand it
Tech bust• All projects must justify expense, security was neglected
4
Brief History of Internet Security, cont.
Today• Little security except governments, banks, and telcos
The Future• Annoying/disruptive malware is for amateurs
• Intellectual property theft is the future
• Cheap and effective industrial espionage
• Database, file server, web-based CMS are next targets
• Popularity of web services (SOAP, ReST) lead to new attack forms
5
Cyber Crime Statistics
Affected 55 million Americans in 2005 (USA Today)
Yearly costs estimates vary wildly• $105 billion (2005, Treasury Department)
• $1 billion (2003, Gartner)
42% of Americans used internet LESS in 2005• Identity theft scares them away (Gartner)
Customers go elsewhere after security breaches (Ponemon)• 70% would leave after two thefts of customer data
• Each lost customer record lost costs on average: $75
6
What Should We Do?
Don’t panic!• 100% security neither possible nor cost-effective
Ultimately, people make your assets secure• Security products (hardware, software) merely assist
• Make your network defensible, not impregnable
Consider security like risk management• Security breaches always possible
• Can lower the odds to zero, but for progressively more money
• Minimize risk, but also keep costs under control
“Security is a process, not a product” – Bruce Schneier
7
Making A Security Plan
1. Determine risks and threats• Who is attacking you, and why?
2. Describe policy required to defend• Who should be allowed to do what?
• Sarbanes Oxley compliance may also require this
3. Analyze vulnerabilities, design countermeasures• Cover all aspects: Protection, Detection, Reaction
8
1) Determine Threats and Risks
Definition of Risk• Probability and effects of a harmful event
• Risk = Threat x Vulnerability x Impact
• Threat: person with skills and motive to leverage vulnerability
Standard Consumers – LOW Risk• Customers, partners, site visitors, Google, competitors
• Using leaked information against you
Contributors – MEDIUM Risk• Typical employees, Site Studio contributors
• Information leakage, bad passwords, malicious content
9
1) Determine Threats and Risks, cont.
Administrators – HIGH Risk• Site maintainer
• Setting security improperly, stealing content, clearing log files
Developers – HIGH Risk• Component, fragment, or portal designers
• Creating insecure code on your system
Unauthorized Users – HIGHEST Risk• Hackers, malware, ex-employees, differs based on organization
Oft cited stat: 70% of attacks are insiders
10
2) Determine Security Policy
Specifics depend on organizational needs
Enterprise network policy• General process for users, administrators, applications
• Most likely you have one already
Content Server specific policies• Who is allowed to view/change each type of content?
• When is content fit for consumption?
• Do contributors need access to scripting languages?
• JSP, ASP, ASP.NET, JavaScript
• Which add-ons and components are secure?
11
3) Vulnerabilities and Countermeasures
Protection, Detection, Reaction
Check vulnerabilities common to all networked systems• Security holes in OS, web server, database
• Bad passwords, eavesdropping, malware
• Evil developers and administrators
• Probably addressed in your current security policy
Vulnerabilities common to web apps• Cross Site Scripting (XSS)
• Information leakage
12
Vulnerabilities in Oracle UCM
Java-based, so “immune” to buffer overflows
Primary vulnerabilities in interfaces• Connection to authentication engine
• Web server, portal server, LDAP
• Database connections (SQL injection)
• Administrative tools can impersonate users
Contributors uploading insecure content• Incorrect security group
• Malicious ASP / JSP / JavaScript / HTML Forms
Insecure components• Especially when they connect to remote systems
13
Protection - High Risk Vulnerabilities
Important regardless of your organization
General network security
Cross site scripting
Malicious content
Content server access port (4444)
14
General Network Security
Keep software/firmware patches up-to-date• 99% of attacks are known exploits (CERT)
Vulnerability analysis tools (Nessus, Metasploit)• Highly flawed, but catches known exploits
Secure network topology• Firewall both external and internal attacks (DMZ)
• Minimize required trust and single points of failure
Restrict physical / shell access
Virus/Trojan scanners at internet gateway• Block malicious emails & web downloads
• Cannot trust users to make secure decisions!
15
Cross-Site Scripting (XSS)
HTML & HTTP flexibility very useful:• Form on my site to submit a Google search
• Click submit, search results displayed in your browser
But could cause security problems:• Click submit, content deleted from your server invisible to you
Attacks can originate from ANYWHERE:• Form submission, URLs, or simple page loading
• Attacks can be JavaScript, or pure HTML
Executed with your browser & security credentials• Hacker tricks you into attacking the server for him
• Advanced attacks with AJAX can hijack your entire browser!
16
Cross-Site Scripting, cont.
Block forms submitted from untrusted sites• Block insecure HTTP referrers
• Referrer can be spoofed, but in this case it’s useful
• Block malicious URLs: EnableSecuredGets=true
• More flags available in the HtmlPostAuthenticator component
Scrub XSS attacks from request data• Use encodeHtml function, and all HtmlDataInput config flags
• Blocks JavaScript in metadata, content, URL parameters
• Vital for anonymous users, might restrict “power users”
• Available in pre-10gr3 as the FilterDataInput component
• Out of the box in 10gr3
17
Malicious Content
Virus scanner on ucm/vault/~temp directory• Checks for malicious / careless contributors
Lock down JSP/ASP contribution to just developers• Lock down JSP with JspEnabledGroups
• Enable ASP at web server only for specific security folders
• http://myhost/stellent/groups/scripts/
• Remove secure data from JSPs/ASPs
• Consumers with access to native file can view the source
Treat malicious HCSP same as malicious HTML• IdocScript limited to page display for security reasons
18
Direct UCM Port Access
Authentication performed outside of UCM• Web server authenticates, passes username to port 4444
• Out of the box, only localhost is trusted
• With dedicated server & trusted admins, fairly safe
Rogue administrator can impersonate users• Content Integration Suite (CIS), RIDC, IdcCommandUX
• SOAP is safer – password authenticated by web server
First: block access with firewalls• Block all but web server and portal server
• Restrict to specific named users if possible
• More safe than just SocketHostAddressSecurityFilter
19
Direct SCS Port Access, cont.
Second: password secure access• Set ProxyPassword to a large random string
• Web server sends secure authentication token:
• User, hashed user password, IP, time, and ProxyPassword
• Blocks connections without token, including some admin tools
• CIS, RIDC, IdcCommandUX
Third: encrypt all communication on port 4444• Uses SSL Certificates
• For J2EE connections only (CIS/CPS/RIDC)
• Use the SecurityProviders component, and the RIDC Developer’s Guide
NOTE: depending on version, 2&3 might not be compatible
20
Protection - Lower Risk Vulnerabilities
Less risky, depending on organization and network
Data leakage
Eavesdropping
Passwords
Hacked database, file server, web server
21
Data Leakage
How much does Google know about you, without even trying?
Simplify the security model• Less accidental information leakage
• Use security groups to secure, not to organize!
• Use Profiles / WebDAV to simplify contribution
Workflows for public facing content
PDF Watermark, Dynamic Watermark• Embed the word CONFIDENTIAL in converted PDFs
• Add date and viewer’s name with Dynamic Watermark
NeedToKnow component• User can view some metadata, but not content
22
Eavesdropping
Network vulnerable to packet sniffers• Spy on downloaded content or passwords
Use SSL (HTTPS) • Protects passwords and content• Important for Basic authentication and cookie login
• NTLM authentication is already “encrypted”
• Drawback: performance suffers
Customizations: use SSL only when needed• Form-based login with cookies from HTTPS page• Set up both HTTP and HTTPS (BrowserUrlPath)
• Use SSL only for contributors / administrators
23
Password Hacking
Cannot trust users with their passwords• 70% give away passwords for chocolate (BBC)• Computers getting faster, passwords still 8 letters• Reused passwords, easily guessed passwords
Use Single Sign On (SSO)• NTLM, Active Directory, LDAP, Netegrity SiteMinder• Eases maintenance of passwords, users, new applications• Drawback: no login prompt can make XSS worse
Don’t allow users to know their full access key!• USB / smart cards with SSL certificates for authentication• Protect certificate with password
• Add biometrics for more security
• In practice, managing such keys is difficult and time consuming• If security protocols are tough, people make “back doors” to do their jobs
24
Hacked Web Server
Web server primary target for unauthorized users• Typically exposed to outside world
• Firewalls help, but don’t stop everything
Add a reverse proxy for extra security layer• Run web server on same machine as UCM
• Connect to web server with reverse proxy outside firewall
• If proxy is hacked, UCM is still safe (for now)
25
Hacked File System
Compromise of UCM machine or file server• If user gains root access, anything is possible
Use NTLM/Kerberos authentication for Database • No need to store JdbcPassword
• SQL Server: Use JTDS driver, manual install
• Oracle: Use version-specific thin driver
Encrypted file system• Secure the vault, weblayout, search, data directories
• Performance loss, might not work with your web server
• Use FileStoreProviders to encrypt just a handful of items
26
Database Connections
SQL Injection attacks• Similar to XSS – malicious SQL through hacked web form
• Can take over database through some web applications
SCS protected by abstraction layer• SQL in predefined queries, strongly typed parameters
SCS validates user-submitted data• Parses dates, numbers, escapes quotes in strings
• Extra parenthesis added if found in query terms
Reduce table-modification rights of Stellent JDBC user• Adding metadata, schema, components requires DBA approval
• User-level activity unaffected
27
Protection - Risks in Customizations
Difficult to measure
Untrusted developer making customizations
How to make secure components
Possible security customizations
28
Malicious Components
Difficult to defend against• Requires complete auditing of code for back-doors
Install “development” servers• Protects content against rogue developers
• Components audited and installed by trusted admin
Block Admin Server component installs• Restrict to Component Wizard from the console
• AllowUpdateComponentConfig, AllowComponentUpload
29
Secure Component Tips
Validate HTML form data in Java• JavaScript is handy, but not trustable
Encode IdocScript variables on pages• IdocScript functions: js, xml, url, encodeHtml
• Prevents introduction of new XSS attacks
Use predefined SQL queries, avoid raw SQL• Predefined queries thwart SQL injection attacks
Validate all code that is sent to a new server, or new process• Check all command-line calls for malicious attacks
• Understand your remote system well enough to stop attacks
30
Secure Component Tips, cont.
Use the checkSecurity action in custom services• Service security flags require dSecurityGroup to be present
• Cannot trust values sent by the user
• First determine dSecurityGroup with SQL query
• Then call checkSecurity action in service
Be paranoid about malicious usage of custom IdocScript• Restrict functionality to read-only actions
• Workflow IdocScript allows write actions in special context
• Condition variable allowWorkflowIdocScript
31
Security Customization Examples
Modify security with Java components• Web security plugins require C++ code
Allow anonymous user to check in content• SecurityFilter component in HowToComponents
• Deny general access, allow it for specific requests
Establish download quotas, or download behavior monitors
Block certain service calls for consumption servers• Block SOAP or WebDAV explicitly
• With firewall, or customization
• Block all services except bare essentials (SS_GET_PAGE)
32
Security Customization Examples, cont.
Custom validation of metadata and content• Content Profile rules, or Java filters (validateStandard)
Java User Providers for authentication• For legacy or atypical authentication protocols
Custom web authentication• Custom authentication filters (SiteMinder, SSL Certificates)
• Custom tokens for flexible SOAP security
• Block login after 3 bad passwords, if not already a rule in SSO system
Custom cookie login• Existing sample version doesn’t have a timeout
• Be sure to hash the username, a timestamp, and a secret
33
Detection – Network Violations
Check logs frequently• Firewall, intrusion detection, web server
• Sometimes only human eyes catch ‘suspicious’ behavior
Syndicate access logs to remote servers• Must compromise several systems to “cover your tracks”
Intrusion detection systems (Snort, some firewalls)• Detects known attacks, generally suspicious packets
34
Detection – UCM Specific
Check SCS web logs• Access violation logged
• Login failures not automatically logged
• Use SSO or custom component
Content Tracker• Tracks popularity – who viewed what?
• Check for allowed – but unusual – activity:
• Downloading data before quitting
• Downloading random highly secure data
• Excellent dual-purpose technology
35
Reaction – Respond to Intrusion
Someone someday will bypass all your security• Reaction depends on business needs
• Block access with firewall? Unplug from network?
• Minimize damage and disruption while you patch
Establish backup and recovery plans• Backup all applications and data
• Most high-availability sites have sufficient plans
Redundancy is primary• System must survive a crash / compromise of one server
• One compromise shouldn’t lead to several
36
Reaction – UCM specific
Block hacked user accounts• Supported by some SSO vendors
• Block the administrator at least (SysadminBlocker)
Make database read-only• Block ‘commit’ rights to Stellent’s JDBC user at the database
• Blocks ALL write activity, possible side effects
• Make Counters table read-only
• Blocks all transactions
37
Conclusions
Oracle UCM less vulnerable than most systems• Smaller target than operating system, database
• Recommendation: minimize XSS and direct port access
• Others countermeasures as needed
Which countermeasures to use?• First, use dual-purpose technology
• SSO, web proxies, Content Tracker
• Second, select those with little cost/drawbacks
• ProxyCredentials, SSL, simplified security model
• Lastly, select those with significant cost/drawbacks
• Scrub JavaScript content, SSL Providers, custom components
38
Now Available…
39
My Company: http://bezzotech.com
My Blog: http://bexhuff.com
My Self: [email protected]
Questions?