1. Worms 1

2. Viruses don’t break into your computer – they are invited by you – They cannot spread unless you run infected application or click on infected attachment – Early viruses spread onto different applications on your computer – Contemporary viruses spread as attachments through E-mail, they will mail themselves to people from your addressbook Worms break into your computer using some vulnerability, install malicious code and move on to other machines – You don’t have to do anything to make them spread 2 Viruses vs. Worms

3. A program that: – Scans network for vulnerable machines – Breaks into machines by exploiting the vulnerability – Installs some piece of malicious code – backdoor, DDoS tool – Moves on Unlike viruses – Worms don’t need any user action to spread – they spread silently and on their own – Worms don’t attach themselves onto other programs – they exist as a separate code in memory Sometimes you may not even know your machine has been infected by a worm 3 What is a Worm?

4. They spread extremely fast They are silent Once they are out, they cannot be recalled They usually install malicious code They clog the network 4 Why Are Worms Dangerous?

5. Robert Morris, a PhD student at Cornell, was interested in network security He created the first worm with a goal to have a program live on the Internet in Nov. 1988 – Worm was supposed only to spread, fairly slowly – It was supposed to take just a little bit of resources so not to draw attention to itself – But things went wrong … Worm was supposed to avoid duplicate copies by asking a computer whether it is infected – To avoid false “yes” answers, it was programmed to duplicate itself every 7 th time it received “yes” answer – This turned out to be too much 5 First Worm Ever – Morris Worm

6. It exploited four vulnerabilities to break in – A bug in sendmail – A bug in finger deamon – A trusted hosts feature (/etc/.rhosts) – Password guessing Worm was replicating at a much faster rate than anticipated At that time Internet was small and homogeneous (SUN and VAX workstations running BSD UNIX) It infected around 6,000 computers, one tenth of then-Internet, in a day 6 First Worm Ever – Morris Worm

7. People quickly devised patches and distributed them (Internet was small then) A week later all systems were patched and worm code was removed from most of them No lasting damage was caused Robert Morris paid $10,000 fine, was placed on probation and did some community work Worm exposed not only vulnerabilities in UNIX but moreover in Internet organization Users didn’t know who to contact and report infection or where to look for patches 7 First Worm Ever – Morris Worm

8. In response to Morris Worm DARPA formed CERT (Computer Emergency Response Team) in November 1988 – Users report incidents and get help in handling them from CERT – CERT publishes security advisory notes informing users of new vulnerabilities that need to be patched and how to patch them – CERT facilitates security discussions and advocates better system management practices 8 First Worm Ever – Morris Worm

9. Spread on July 12 and 19, 2001 Exploited a vulnerability in Microsoft Internet Information Server that allows attacker to get full access to the machine (turned on by default) Two variants – both probed random machines, one with static seed for RNG, another with random seed for RNG (CRv2) CRv2 infected more than 359,000 computers in less than 14 hours – It doubled in size every 37 minutes – At the peak of infection more than 2,000 hosts were infected each minute 9 Code Red

10. 10 Code Red v2

11. 43% of infected machines were in US 47% of infected machines were home computers Worm was programmed to stop spreading at midnight, then attack www1.whitehouse.gov – It had hardcoded IP address so White House was able to thwart the attack by simply changing the IP address-to-name mapping Estimated damage ~2.6 billion 11 Code Red v2

12. Spread on January 25, 2003 The fastest computer worm in history – It doubled in size every 8.5 seconds. – It infected more than 90% of vulnerable hosts within 10 minutes – It infected 75,000 hosts overall Exploited buffer overflow vulnerability in Microsoft SQL server, discovered 6 months earlier 12 Sapphire/Slammer Worm

13. No malicious payload The aggressive spread had severe consequences – Created DoS effect – It disrupted backbone operation – Airline flights were canceled – Some ATM machines failed 13 Sapphire/Slammer Worm

14. 14 Sapphire/Slammer Worm

15. Both Slammer and Code Red 2 use random scanning o Code Red uses multiple threads that invoke TCP connection establishment through 3-way handshake – must wait for the other party to reply or for TCP timeout to expire o Slammer packs its code in single UDP packet – speed is limited by how many UDP packets can a machine send o Could we do the same trick with Code Red? Slammer authors tried to use linear congruential generators to generate random addresses for scanning, but programmed it wrong 15 Why Was Slammer So Fast?

16. 43% of infected machines were in US 59% of infected machines were home computers Response was fast – after an hour sites started filtering packets for SQL server port 16 Sapphire/Slammer Worm

17. 17 BGP Impact of Slammer Worm

18. 18 Stuxnet Worm Discovered in June/July 2010 Targets industrial equipment Uses Windows vulnerabilities (known and new) to break in Installs PLC (Programmable Logic Controller) rootkit and reprograms PLC – Without physical schematic it is impossible to tell what’s the ultimate effect Spread via USB drives Updates itself either by reporting to server or by exchanging code with new copy of the worm

19. Many worms use random scanning This works well only if machines have very good RNGs with different seeds Getting large initial population represents a problem – Then the infection rate skyrockets – The infection eventually reaches saturation since all machines are probing same addresses 19 Scanning Strategies “Warhol Worms: The Potential for Very Fast Internet Plagues”, Nicholas C Weaver

20. 20 Random Scanning

21. Worm can get large initial population with hitlist scanning Assemble a list of potentially vulnerable machines prior to releasing the worm – a hitlist – E.g., through a slow scan When the scan finds a vulnerable machine, hitlist is divided in half and one half is communicated to this machine upon infection – This guarantees very fast spread – under one minute! 21 Scanning Strategies

22. 22 Hitlist Scanning

23. Worm can get prevent die-out in the end with permutation scanning All machines share a common pseudorandom permutation of IP address space Machines that are infected continue scanning just after their point in the permutation – If they encounter already infected machine they will continue from a random point Partitioned permutation is the combination of permutation and hitlist scanning – In the beginning permutation space is halved, later scanning is simple permutation scan 23 Scanning Strategies

24. 24 Permutation Scanning

25. Worm can get behind the firewall, or notice the die-out and then switch to subnet scanning Goes sequentially through subnet address space, trying every address 25 Scanning Strategies

26. Several ways to download malicious code – From a central server – From the machine that performed infection – Send it along with the exploit in a single packet 26 Infection Strategies

27. Three factors define worm spread: – Size of vulnerable population Prevention – patch vulnerabilities, increase heterogeneity – Rate of infection (scanning and propagation strategy) Deploy firewalls Distribute worm signatures – Length of infectious period Patch vulnerabilities after the outbreak Worm Defense

28. This depends on several factors: – Reaction time – Containment strategy – address blacklisting and content filtering – Deployment scenario – where is response deployed Evaluate effect of containment 24 hours after the onset How Well Can Containment Do? “Internet Quarantine: Requirements for Containing Self-Propagating Code”, Proceedings of INFOCOM 2003, D. Moore, C. Shannon, G. Voelker, S. Savage

29. How Well Can Containment Do? Code Red Idealized deployment: everyone deploys defenses after given period

30. How Well Can Containment Do? Depending on Worm Aggressiveness Idealized deployment: everyone deploys defenses after given period

31. How Well Can Containment Do? Depending on Deployment Pattern

32. Reaction time needs to be within minutes, if not seconds We need to use content filtering We need to have extensive deployment on key points in the Internet How Well Can Containment Do?

33. Monitor outgoing connection attempts to new hosts When rate exceeds 5 per second, put the remaining requests in a queue When number of requests in a queue exceeds 100 stop all communication Detecting and Stopping Worm Spread “Implementing and testing a virus throttle”, Proceedings of Usenix Security Symposium 2003, J. Twycross, M. Williamson

34. Detecting and Stopping Worm Spread

36. Organizations share alerts and worm signatures with their “friends” – Severity of alerts is increased as more infection attempts are detected – Each host has a severity threshold after which it deploys response Alerts spread just like worm does – Must be faster to overtake worm spread – After some time of no new infection detections, alerts will be removed Cooperative Strategies for Worm Defense “Cooperative Response Strategies for Large-Scale Attack Mitigation”, Proceedings of DISCEX 2003, D. Norjiri, J. Rowe, K. Levitt

37. As number of friends increases, response is faster Propagating false alarms is a problem Cooperative Strategies for Worm Defense

38. Early detection would give time to react until the infection has spread The goal of this paper is to devise techniques that detect new worms as they just start spreading Monitoring: – Monitor and collect worm scan traffic – Observation data is very noisy so we have to filter new scans from Old worms’ scans Port scans by hacking toolkits Early Worm Detection C. C. Zou, W. Gong, D. Towsley, and L. Gao. "The Monitoring and Early Detection of Internet Worms," IEEE/ACM Transactions on Networking.

39. Detection: – Traditional anomaly detection: threshold-based Check traffic burst (short-term or long-term). Difficulties: False alarm rate – “Trend Detection” Measure number of infected hosts and use it to detect worm exponential growth trend at the beginning Early Worm Detection

40. Worms uniformly scan the Internet – No hitlists but subnet scanning is allowed Address space scanned is IPv4 Assumptions

41. Simple epidemic model: Worm Propagation Model Detect worm here. Should have exp. spread

42. Monitoring System

43. Provides comprehensive observation data on a worm’s activities for the early detection of the worm Consists of : – Malware Warning Center (MWC) – Distributed monitors Ingress scan monitors – monitor incoming traffic going to unused addresses Egress scan monitors – monitor outgoing traffic Monitoring System

44. Ingress monitors collect: – Number of scans received in an interval – IP addresses of infected hosts that have sent scans to the monitors Egress monitors collect: – Average worm scan rate Malware Warning Center (MWC) monitors: – Worm’s average scan rate – Total number of scans monitored – Number of infected hosts observed Monitoring System

45. MWC collects and aggregates reports from distributed monitors If total number of scans is over a threshold for several consecutive intervals, MWC activates the Kalman filter and begins to test the hypothesis that the number of infected hosts follows exponential distribution Worm Detection

46. Population: N=360,000, Infection rate:  = 1.8/hour, Scan rate  = 358/min, Initially infected: I 0 =10 Monitored IP space 2 20, Monitoring interval:  = 1 minute Code Red Simulation Infected hosts  estimation

47. Population: N=100,000 Scan rate  = 4000/sec, Initially infected: I 0 =10 Monitored IP space 2 20, Monitoring interval:  = 1 second Slammer Simulation Infected hosts  estimation

48. Dynamic Quarantine Worms spread very fast (minutes, seconds) – Need automatic mitigation If this is a new worm, no signature exists – Must apply behaviour-based anomaly detection – But this has a false-positive problem! We don’t want to drop legitimate connections! Dynamic quarantine – “Assume guilty until proven innocent” – Forbid access to suspicious hosts for a short time – This significantly slows down the worm spread C. C. Zou, W. Gong, and D. Towsley. "Worm Propagation Modeling and Analysis under Dynamic Quarantine Defense," ACM CCS Workshop on Rapid Malcode (WORM'03),

49. Dynamic Quarantine Behavior-based anomaly detection can point out suspicious hosts – Need a technique that slows down worm spread but doesn’t hurt legitimate traffic much – “Assume guilty until proven innocent” technique will briefly drop all outgoing connection attempts (for a specific service) from a suspicious host – After a while just assume that host is healthy, even if not proven so – This should slow down worms but cause only transient interruption of legitimate traffic

50. Dynamic Quarantine Assume we have some anomaly detection program that flags a host as suspicious – Quarantine this host – Release it after time T – The host may be quarantined multiple times if the anomaly detection raises an alarm – Since this doesn’t affect healthy hosts’ operation a lot we can have more sensitive anomaly detection technique

51. Dynamic Quarantine An infectious host is quarantined after time units A susceptible host is falsely quarantined after time units Quarantine time is T, after that we release the host A few new categories: – Quarantined infectious R(t) – Quarantined susceptible Q(t)

52. Slammer With DQ

53. DQ With Large T? T=10sec T=30sec

54. DQ And Patching?

55. Patch Only Quarantined Hosts Cleaning I(t) Cleaning R(t)

56. DOMINO The goal is to build an overlay network so that nodes cooperatively detect intrusion activity – Cooperation reduces the number of false positives Overlay can be used for worm detection Main feature are active-sink nodes that detect traffic to unused IP addresses The reaction is to build blacklists of infected nodes V. Yegneswaran, P. Barford, S. Jha, “Global Intrusion Detection in the DOMINO Overlay System,” NDSS 2004

57. DOMINO Architecture

58. Axis nodes collect, aggregate and share data – Nodes in large, trustworthy ISPs – Each node maintains a NIDS and an active sink over large portion of unused IP space Access points grant access to axis nodes after thorough administrative checks Satellite nodes form trees below an axis node, collect information, deliver it to axis nodes and pull relevant information Terrestrial nodes supply daily summaries of port scan data

59. Information Sharing Every axis node maintains a global and local view of intrusion activity Periodically a node receives summaries from peers which are used to update global view – List of worst offenders grouped per port – Lists of top scanned ports RSA is used to authenticate nodes and signed SHA digests are used to ensure message integrity and authenticity

60. How Many Nodes We Need? 40 for port summaries 20 for worst offender list

61. How Frequent Info Exchange? Staleness doesn’t matter much but more frequent lists are better to catch worst offenders

62. How Long Blacklists? About 1000 IPs are enough

63. How Close Monitoring Nodes? Blacklists in same /16 space are similar  satellites in /16 space should be grouped under the same axis node and sets of /16 spaces should be randomly distributed among different axis nodes

64. Automatic Worm Signatures Focus on TCP worms that propagate via scanning Idea: vulnerability exploit is not easily mutable so worm packets should have some common signature Step 1: Select suspicious TCP flows using heuristics Step 2: Generate signatures using content prevalence analysis Kim, H.-A. and Karp, B., Autograph: Toward Automated, Distributed Worm Signature Detection, in the Proceedings of the 13th Usenix Security Symposium (Security 2004), San Diego, CA, August, 2004.

65. Suspicious Flows Detect scanners as hosts that make many unsuccessful connection attempts (>2) Select their successful flows as suspicious Build suspicious flow pool – When there’s enough flows inside trigger signature generation step

66. Signature Generation Use most frequent byte sequences across flows as the signature Naïve techniques fail at byte insertion, deletion, reordering Content-based payload partitioning (COPP) – Partition if Rabin fingerprint of a sliding window matches breakmark = content blocks – Configurable parameters: window size, breakmark – Analyze which content blocks appear most frequently and what is the smallest set of those that covers most/all samples in suspicious flow pool

67. How Well Does it Work? Tested on traces of HTTP traffic interlaced with known worms For large block sizes and large coverage of suspicious flow pool (90-95%) Autograph performs very well – Small false positives and false negatives

68. Distributed Autograph Would detect more scanners Would produce more data for suspicious flow pool – Reduce false positives and false negatives

69. Automatic Signatures (approach 2) Detect content prevalence – Some content may vary but some portion of worm remains invariant Detect address dispersion – Same content will be sent from many hosts to many destinations Challenge: how to detect these efficiently (low cost = fast operation) S.Singh, C. Estan, G. Varghese and S. Savage “Automated Worm Fingerprinting,” OSDI 2004

70. Content Prevalence Detection Hash content + port + proto and use this as key to a table where counters are kept – Content hash is calculated over overlapping blocks of fixed size – Use Rabin fingerprint as hash function – Autograph calculates Rabin fingerprint over variable-length blocks that are non-overlapping – Rabin fingerprint is a hash function that is efficient to recalculate if a portion of the input changes

71. Address Dispersion Detection Remembering sources and destinations for each content would require too much memory Scaled bitmap: – Sample down input space, e.g., hash into values 0- 63 but only remember those values that hash into 0-31 – Set the bit for the output value (out of 32 bits) – Increase sampling-down factor each time bitmap is full = constant space, flexible counting

72. How Well Does This Work? Implemented and deployed at UCSD network

73. How Well Does This Work? Some false positives – Spam, common HTTP protocol headers.. (easily whitelisted) – Popular BitTorrent files (not easily whitelisted) No false negatives – Detected each worm outbreak reported in news – Cross-checked with Snort’s signature detection

74. Polymorphic Worm Signatures Insight: multiple invariant substrings must be present in all variants of the worm for the exploit to work – Protocol framing (force the vulnerable code down the path where the vulnerability exists) – Return address Substrings not enough = too short Signature: multiple disjoint byte strings – Conjunction of byte strings – Token subsequences (must appear in order) – Bayes-scored substrings (score + threshold) J. Newsome, B. Karp and D. Song, “Polygraph: Automatically Generating Signatures for Polymorphic Worms,” IEEE Security and Privacy Symposium, 2005

75. Worm Code Structure Invariant bytes: any change makes the worm fail Wildcard bytes: any change has no effect Code bytes: Can be changed using some polymorphic technique and worm will still work – E.g., encryption

76. Polygraph Architecture All traffic is seen, some is identified as part of suspicious flows and sent to suspicious traffic pool – May contain some good traffic – May contain multiple worms Rest of traffic is sent to good traffic pool Algorithm makes a single pass over pools and generates signatures

77. Signature Detection Extract tokens (variable length) that occur in at least K samples – Conjuction signature is this set of tokens – To find token-subsequence signatures samples in the pool are aligned in different ways (shifted left or right) so that the maximum-length subsequences are identified – Contiguous tokens are preferred – For Bayes signatures for each token a probability is computed that it is contained by a good or a suspicious flow – use this as a score – Set high value of threshold to avoid false positives

78. How Well Does This Work? Legitimate traffic traces: HTTP and DNS – Good traffic pool – Some of this traffic mixed with worm traffic to model imperfect separation Worm traffic: Ideally-polymorphic worms generated from 3 known exploits Various tests conducted

79. How Well Does This Work? When compared with single signature (longest substring) detection, all proposed signatures result in lower false positive rates – False negative rate is always zero if the suspicious pool has at least three samples If some good traffic ends up in suspicious pool – False negative rate is still low – False positive rate is low until noise gets too big If there are multiple worms in suspicious pool and noise – False positives and false negatives are still low

80. Borrowed from Brent ByungHoon Kang, GMU

81. A Network of Compromised Computers on the Internet IP locations of the Waledac botnet. Borrowed from Brent ByungHoon Kang, GMU

82. Botnets Networks of compromised machines under the control of hacker, “bot-master” Used for a variety of malicious purposes: Sending Spam/Phishing Emails Launching Denial of Service attacks Hosting Servers (e.g., Malware download site) Proxying Services (e.g., FastFlux network) Information Harvesting (credit card, bank credentials, passwords, sensitive data.) Borrowed from Brent ByungHoon Kang, GMU

83. Botnet with Central Control Server After resolving the IP address for the IRC server, bot-infected machines CONNECT to the server, JOIN a channel, then wait for commands. Borrowed from Brent ByungHoon Kang, GMU

84. Botnet with Central Control Server The botmaster sends a command to the channel. This will tell the bots to perform an action. Borrowed from Brent ByungHoon Kang, GMU

85. Botnet with Central Control Server The IRC server sends (broadcasts) the message to bots listening on the channel. Borrowed from Brent ByungHoon Kang, GMU

86. Botnet with Central Control Server The bots perform the command. In this example: attacking / scanning CNN.COM. Borrowed from Brent ByungHoon Kang, GMU

87. Botnet Sophistication Fueled by Underground Economy Unfortunately, the detection, analysis and mitigation of botnets has proven to be quite challenging Supported by a thriving underground economy – Professional quality sophistication in creating malware codes – Highly adaptive to existing mitigation efforts such as taking down of central control server. 87 Borrowed from Brent ByungHoon Kang, GMU

88. Emerging Decentralized Peer to Peer Multi-layered Botnets Traditional botnet communication – Central IRC server for Command & Control (C&C) – Single point of mitigation: C&C Server can be taken down or blacklisted Botnets with peer to peer C&C – No single point of failure. – E.g., Waldedac, Storm, and Nugache Multi-layered architecture to obfuscate and hide control servers in upper tiers. Borrowed from Brent ByungHoon Kang, GMU

89. Expected Use of DHT P2P Network Publish and Search  Botmaster publishes commands under the key.  Bots are searching for this key periodically  Bots download the commands =>Asynchronous C&C Borrowed from Brent ByungHoon Kang, GMU

90. Multi-Layered Command and Control Architecture Through P2P  Each Supernode (server) publishes its location (IP address) under the key 1 and key 2  Subcontrollers search for key 1  Subnodes (workers) search for key 2 to open connection to the Supernodes => Synchronous C&C  Borrowed from Brent ByungHoon Kang, GMU

91. Current Botnet Defenses Virus Scanner at Local Host – Polymorphic binaries against signature scanning – Not installed even though it is almost free – Rootkit Network Intrusion Detection Systems – Keeping states for network flows – Deep packet inspection is expensive – Deployed at LAN, and not scalable to ISP-level – Requires Well-Trained Net-Security SysAdmin Borrowed from Brent ByungHoon Kang, GMU

92. 92 Conficker infections are still increasing after one year!!! There are millions of computers on the Internet that do not have virus scanner nor IDS There are millions of computers on the Internet that do not have virus scanner nor IDS Borrowed from Brent ByungHoon Kang, GMU

93. Botnet Enumeration Approach Used for spam blocking, firewall configuration, DNS rewriting, and alerting sys-admins regarding local infections. Fundamentally differs from existing Intrusion Detection System (IDS) approaches – IDS protects local hosts within its perimeter (LAN) – An enumerator would identify both local as well as remote infections Identifying remote infections is crucial – There are numerous computers on the Internet that are not under the protection of IDS-based systems. 93 Borrowed from Brent ByungHoon Kang, GMU

94. How to Enumerate Botnet Need to know the method and protocols for how a bot communicates with its peers Using sand-box technique – Run bot binary in a controlled environment – Network behaviors are captured/analyzed Investigating the binary code itself – Reversing the binary into high level codes – C&C Protocol knowledge and operation details can be accurately obtained Borrowed from Brent ByungHoon Kang, GMU

95. Simple Crawler Approach Given network protocol knowledge, crawlers: 1.collect list of initial bootstrap peers into queue 2.choose a peer node from the queue 3.send to the node look-up or get-peer requests 4.add newly discovered peers to the queue 5.repeat 2-5 until no more peer to be contacted Can’t enumerate a node behind NAT/Firewall Would miss bot-infected hosts at home/office! Borrowed from Brent ByungHoon Kang, GMU

96. Passive P2P Monitor (PPM) Given P2P protocol knowledge that bot uses A collection of “routing-only” nodes that – Act as peer in the P2P network, but – Controlled by us, the defender PPM nodes can observe the traffic from the peer infected hosts PPM node can be contacted by the infected hosts behind NAT/Firewall Borrowed from Brent ByungHoon Kang, GMU

97. Crawler and Passive P2P Monitor (PPM) Crawler PPM Borrowed from Brent ByungHoon Kang, GMU

98. Crawler vs. PPM: # of IPs found Borrowed from Brent ByungHoon Kang, GMU

99. Botnet Enumeration Challenges DHCP NATs Non-uniform bot distribution Churn Most estimates put size of largest botnets at tens of millions of bots – Actual size may be much smaller if we account for all of the above

100. Fast Flux Botnets use a lot of newly-created domains for phishing and malware delivery Fast flux: changing name-to-IP mapping very quickly, using various IPs to thwart defense attempts to bring down botnet Single-flux: changing name-to-IP mapping for individual machines, e.g., a Web server Double-flux: changing name-to-IP mapping for DNS nameserver too Proxies on compromised nodes fetch content from backend servers

101. Fast Flux Advantages for the attacker: – Simplicity: only one back end server is needed to deliver content – Layers of protection through disposable proxy nodes – Very resilient to attempts for takedown

102. Fast Flux Detection Look for domain names where mapping to IP changes often – May be due to load balancing – May have other (non-botnet) cause, e.g., adult content delivery – Easy to fabricate domain names Look for DNS records with short-lived domain names, with lots of A records, lots of NS records and diverse IP addresses (wrt AS and network access type) Look for proxy nodes by poking them

103. Poking Botnets is Dangerous They have been known to fight back – DDoS IPs that poke them (even if low workers are scanned) They have been known to fabricate data for honeynets – Honeynet is a network of computers that sits in otherwise unused (dark) address space and is meant to be compromised by attackers

104. Researchers subverted a botnet’s command and control infrastructure (proxy bots) – Modified its spam messages to point to the Web server under researcher control That server mimicked the original Web page from the spam emails – A pharmacy site – A greeting card download site Botnets Fun Facts: ROI for Attackers "Spamalytics: An Empirical Analysis of Spam Marketing Conversion” C. Kanich, C. Kreibich, K. Levchenko, B. Enright, G. Voelker, V. Paxson, and S. Savage, ACM CCS 2009

105. How many spam emails reach recipients: open a few email accounts themselves and append them to email delivery lists in spam messages How many emails result in Web page visits – Must filter out defense accesses How many users actually buy advertised products or download software – No “sale” is finalized Ethical issues abound What Is ROI for Attackers "Spamalytics: An Empirical Analysis of Spam Marketing Conversion” C. Kanich, C. Kreibich, K. Levchenko, B. Enright, G. Voelker, V. Paxson, and S. Savage, ACM CCS 2009

106. Most-targeted E-mail Domains "Spamalytics: An Empirical Analysis of Spam Marketing Conversion” C. Kanich, C. Kreibich, K. Levchenko, B. Enright, G. Voelker, V. Paxson, and S. Savage, ACM CCS 2009

107. Spam Conversion Pipeline "Spamalytics: An Empirical Analysis of Spam Marketing Conversion” C. Kanich, C. Kreibich, K. Levchenko, B. Enright, G. Voelker, V. Paxson, and S. Savage, ACM CCS 2009

108. Spam Conversion Pipeline "Spamalytics: An Empirical Analysis of Spam Marketing Conversion” C. Kanich, C. Kreibich, K. Levchenko, B. Enright, G. Voelker, V. Paxson, and S. Savage, ACM CCS 2009

109. Spam Filter Misses "Spamalytics: An Empirical Analysis of Spam Marketing Conversion” C. Kanich, C. Kreibich, K. Levchenko, B. Enright, G. Voelker, V. Paxson, and S. Savage, ACM CCS 2009

110. For More on Botnets http://www.shadowserver.org http://www.honeynet.org/papers/bots/ http://www.honeynet.org/papers/ff