Introduction
There are only two types of companies: those that have been hacked, and those that will be. Even that is merging into one category: those that have been hacked and will be again.
Robert Mueller FBI Director
Information Security Goals
- C-I-A triad
- Confidentiality
- Unauthorized disclosure of information
- Integrity
- Unauthorized modification of information
- Availability
- Unauthorized withholding of information or resources
- Confidentiality
- Others
- Privacy
- Authenticity
- Non-repudiation
- Accountability
- Auditability
Attack
Attackers and Motivations
- Script-kiddies
- Motivated by curiosity
- Cybercriminals
- Motivated by profit
- Typical demographics: east European, Brazilian
- Nation-state hackers
- Motivated by power
- Typical demographics: east Asian, middle eastern
- Hacktivists
- Motivated by ideology
- Typical demographics: north American, western European
- Cyber-mercenaries
- Hired by to attack
- Insiders
- Motivated by disgruntlement
Vulnerabilities
- Backdoors
- Kleptographic attack
- Rootkit
- Denial of Service
- Resource exhaustion
- Attack amplifiers (e.g., poorly designed FTP, DNS)
- Application or OS exploit
- Eavesdropping
- Listening to private communication on network
- Monitoring hardware electro-magnetic transmissions
- Exploits
- Gain control of a computer system, allow privilege escalation, or denial of service attack
- Used in Trojan horses, viruses
- Social Engineering
- Humans: the weakest link in security
Attack Categories
- Probe
- Information gathering (1:1, 1:m, m:1, m:n modes)
- IPSweep, portsweep, nmap, etc.
- Denial of Service (DoS)
- TCP SYN flood, Ping of Death, smurf, neptune, etc.
- Remote to Local attacks (R2L)
- Brute force/Dictionary attack, buffer overflow, unverified input attacks
- Social engineering, Trojans
- User to Root attacks (U2R)
- Buffer overflow, rootkit, etc.
- Infections
- Trojans/worms/viruses
- Spreading attacks
Basic Attack Steps
- Prepare
- Gather info: Valid IP addresses & ports, OS, software type & version
- Exploit
- Leave behind
- Backdoors
- Clean up
- Restart crashed daemons, clean registry/log files
- Variable order and duration
- Attacker's skill level
- Type of vulnerability to exploit
- Prior knowledge
- Starting location of attacker
Attack Tools
- Information gathering
- Sniffing: capture packets traversing network
- Tcpdump, Ethereal, Gulp, Net2pcap, Dsniff, etc.
- Network mapping/scanning/fingerprinting: hosts/IPs/ports, protocol details
- Nmap, Amap, Vmap, Ttlscan, P0f, Xprobe, Queso, etc.
- Sniffing: capture packets traversing network
- Attack launching
- Trojans
- Danger, NukeNabbler, AIMSpy, NetSpy, etc.
- DoS attacks
- Targa, Burbonic, HOIC, LOIC, etc.
- Packet forging tools
- Packeth, Packit, Packet Excalibur, Nemesis, Tcpinject, Libnet, SendIP, etc.
- Application layer tools
- Code Red Worm, Nimda Worm, AppDDoS, RefRef, etc.
- User attack tools
- Ntfsdos, Yaga, etc.
- Trojans
Advance Persistent Threats: APT
- Targeted attack against a high-value asset
- Low and slow
- Avoid alerts
- Use stolen user credentials
- Zero-day exploits
- Low profile in network
- Slow progress: Operating over months or years
- Beyond limited correlation time windows of today's IDSs
- Multi-stage
- Exploitation
- Command and control
- Lateral movement
- Breach
- Typical Goals
- Steal intellectual property (IP)
- Gain access to sensitive customer data
- Access strategic business information
- Financial gain, embarrassment, blackmail, data poisoning, illegal insider trading, disrupting organization's business
- Attackers
- Well-funded
- Highly skilled
- Motivated
- Targeted on specific data from specific organization
Defense
Detective Security
- 1st Generation: Intrusion detection systems (IDS)
- 100% protection/prevention impossible
- Layered security
- 2nd Generation: Security information and event management (SIEM)
- Correlate alerts from different intrusion detection sensors
- Present actionable information to security analyst
- 3rd Generation: Big Data analytics for security
- Contextual security intelligence
- Long-term correlations
Risk Management: Controls
- Administrative
- Policies, guidelines
- Password policies
- Payment Card Industry Data Security Standard (PCIDSS)
- Principle of least privilege
- Policies, guidelines
- Physical
- Doors, locks, etc.
- Principle of separation of duties
- Logical
- Use software and data
Preventive Measures
- Protocols
- Secure Socket Layer (SSL): source authentication
- Host-based protections
- Secure operating systems, Patching
- Access control
- Identification: username
- Authentication: Something you know/have/are
- Authorization: File permissions, Kerberos, Need-to-know principle
- Firewalls
- Control inter-network traffic (e.g., from/to internet)
- Security by design
- Principle of least privilege, Code reviews, Unit testing, Defense in depth
- Secure coding
- Buffer overflows, Format string vulnerabilities, Code/Command injection
Defense in Depth
- Layered approach
- Separate systems into network sections
- Place firewalls at section boundaries
- Border router between ISP and firewall to filter traffic
- Switches on each section to make sniffing less effective
- Encryption
- Last layer of defense
- Detection
Reactive Defense
- Examples
- Antivirus signatures for known malicious executables
- Email filters for unwanted messages
- Web filters for compromised websites
- Sandboxes for malicious behaviors
- Median detection time between intrusion/breach to awareness of it: 300-400+ days
- Duration of zero-day attacks
- 19 days to 30 months
- Median of 8 months, Average of 10 months
- 61% of attacks discovered by a third party
- Businesses reluctant to disclose their breaches
- Only 2%-30% do
- Porous perimeter
- Cloud applications
- Mobile/BYOD
- Partner businesses
Data Mining in Security
Why Big Data
- Attack landscape
- Attacks increasingly more sophisticated
- Attacking constantly getting easier
- Required attacker knowledge going down
- Quality of attack tools increasing
- Highly motivated attackers
- Attacker needs to succeed only once, defense needs to be right every single time
- Attack mechanisms constantly evolving/mutating, current detection techniques failing
- Polymorphic malwares
- Zero-day attacks
- APTs
- Network perimeter dissolving
- Mobile/BYOD
- Cloud
- Big Data technology enable storage and analysis of higher volumes & more types of data
- 2010 Verizon data breach investigation
- In 86% of cases of breach, evidence was in the logs
- Detection mechanisms failed to raise alerts
- How do we make sense of the data?
Explosion of Malware
- 403 million new variants of malware created in 2011
- 100,000 unique malware samples collected daily by McAfee in 2012, Q1
- More than 100 million samples in McAfee's malware signature database by 2012 Q3
- Practically impossible to keep up with signatures
Detection Taxonomy
Information source
- Host-based
- system calls, system logs
- Network-based
- Wireless Network
- Application logs
- DB logs, web logs
- IDS sensor alerts
- Lower level sensor alarms
Analysis strategy
- Misuse detection
- Premise
- Knowledge of attack patterns provided by human experts
- Signature matching
- Data mining using labeled data sets
- Benefit
- High accuracy in detecting known attacks
- Drawbacks
- Ineffective against novel attacks
- Signatures need updates with each new discovered attack
- Premise
- Anomaly detection
- Premise
- Build profiles of normal behavior (users, hosts, networks)
- Detect deviations from normal profiles
- Benefit
- Detect novel attacks
- Drawback
- Possible high false alarm rate
- Premise
Time aspects
- Real-time
- Analyze live data (e.g., session data)
- Raise alert immediately if attack detected
- Offline
- Analyze data offline
- Useful for forensics
Activeness
- Passive reaction
- Only generate alarms
- Benefit: Human in the loop
- Drawback: Alert may go unnoticed
- Example: Target breach
- Active response
- Corrective response (e.g., reconfigure firewalls)
- Proactive (e.g., log out attacker)
- Benefit: Speed
- Drawback: May turn into DoS attack against
Continuality
- Continuous monitoring
- Continuous real-time analysis
- Collect information about actions immediately
- Higher deployment effort
- Periodic analysis
- Take periodic snapshots of the environment
- Lower security: exploiting the window of opportunity between two snapshots
References
- CS 259D Lecture 1