The proposal for "CRI-Telescope: A Real-time Lens into Dark Address Space of the Internet" is shown below. The CRI-Telescope proposal in PDF is also available.
Contents1 Motivation and Goals
2 Current status of data collection
2.1 Existing telescope instrumentation
2.2 Reporting software
2.3 Prior data releases
2.4 Prior research and education contributions from telescope data
3 Proposed Work
3.1 Task 1: Enhance tools for telescope data analysis and visualization
3.1.1 Network-layer analysis
3.1.2 Transport-layer analysis
3.1.3 Application-layer analysis
3.2 Task 2: Deploy and refine hardware and software for real-time sharing of telescope data
3.2.2 User support
3.3 Task 3: Community Development
3.3.1 Data-sharing policy framework
3.3.2 Workshop on telescope measurements and data
4 Research and education opportunities enabled
4.1 Why CAIDA is the most appropriate team for this project
5 Broader Impact Activities
5.1 Dissemination of results
5.2 Integration of research and education
5.3 Supporting Diversity with CAIDA Activities
1 Motivation and GoalsIn the last decade, network telescopes have been used to observe Internet "background radiation", i.e. unsolicited traffic sent to unassigned address space ("darkspace") . The routing system carries the traffic to darkspace because its address is being announced globally, but there is no response back to the traffic sources since there are no hosts in darkspace. Observing such "one-way" traffic allows visibility into a wide range of security-related events, including scanning of address space by hackers looking for vulnerable targets, backscatter from denial-of-service attacks using random spoofed source addresses, the automated spread of malware such as worms or viruses, and various misconfigurations (e.g., mistyping an IP address). The observed packets represent mostly failed attempts to open connections, or other malware-related behavior. In the last two years this type of traffic has significantly increased due to botnet-related activities such as Conficker's scanning and p2p signaling [2,3]. This year we have also seen a surprising increase in UDP traffic carrying payload that matches the signaling behavior of popular p2p file-sharing software . We propose to expand our telescope instrumentation to enable researchers to exploit this unique global data source to improve our understanding of security-related events such as large-scale attacks and malware spread, and provide researchers and educators opportunities for real-time and historical analyses otherwise inaccessible, and at relatively little cost. Three pervasive challenges in network traffic research, including on the telescope, guide our proposed expansion: collection and storage, efficient curation, and sharing large volumes of data. The volume of data captured by the telescope is expensive to store, limiting the number of researchers who can realistically download and process data sets. The situation is worse during malicious activity outbreaks when the data volumes increase sharply, yet rapid analysis and response are necessary1. Perhaps the most challenging obstacles to sharing any kind of Internet traffic data (even data to unused addresses!) are the privacy and security concerns. Viruses and worms may involve the installation of backdoors that provide unfettered access to infected computers, and telescope data could advertise these especially vulnerable machines. Yet ultimately, the speed, scope, and strength of today's automated malicious software demand effective real-time sources of data that can match the dynamics of the threat. Studying a worm in situ requires real-time traffic access, including raw victim host IP addresses and payload data. We propose to deploy and evaluate an innovative shift in network monitoring that explicitly addresses all three challenges: implement real-time sharing of network traffic data, in a way that maximizes research and education utility while protecting user privacy. We will use a set of attributes we developed for real-time classification of traffic into known types (e.g., DoS attacks, vulnerability scans, etc.) to refine our methods and tools for early detection of meaningful changes in Internet background radiation. Improved reporting of statistics and event triggers and notifications will help researchers understand the macroscopic dynamics of the traffic and draw their attention to aspects of the traffic they may want to study while the event is still happening. To enable such responsiveness, we will enhance the infrastructure for telescope data collection and storage to allow vetted researchers to run analysis programs approximately one hour after data collection. We will support safe and ethical data sharing via our Privacy-Sensitive Sharing Framework (PS2) which integrates privacy-enhancing technology with a policy framework using proven and standard privacy principles and obligations of data seekers and data providers. We will apply our PS2 model to the UCSD network telescope data and evaluate its applicability to other sources of data. To our knowledge, such advanced model of network traffic data-sharing has never been tried before in the academic community, and there are logistic, policy, as well as technical challenges to overcome. But the outcomes we seek - more effective traffic monitoring instrumentation, and privacy frameworks to support sharing the resulting data - can help transform the relatively siloed, below-the-radar data sharing practices of the network and security research community into a more reputable and pervasive scientific discipline. Section 2 describes our current data collection infrastructure and methodology for processing gigabytes of data from the UCSD Network Telescope and prior results. In Section 3, we propose to expand: (1) the technological capability of our darkspace (telescope) instrumentation, including a traffic classification and analysis methodology that will facilitate detection of changes in the nature of Internet background radiation over both long and short time scales; (2) hardware and software support real-time sharing of telescope data; (3) access and usability of gathered data through community development activities, including a new data-sharing policy framework that effectively manages privacy risks, and workshop for researchers interested in using our telescope data, improving telescopes as scientific instrumentation, or sharing their own telescope or other types of data in real-time.
2 Current status of data collection
2.1 Existing telescope instrumentationA telescope's resolution depends on the size of its address base; a telescope using a /16 address prefix will see more packets than one using a /24 prefix. The UCSD network telescope  uses a /8 mostly "dark" (unassigned) network prefix which corresponds to 1/256th of the total IPv4 address space and has only a few assigned addresses. We separate the legitimate traffic destined to those few reachable IP addresses, and monitor only the traffic destined to the empty address space. Therefore, if a host sends packets to uniformly random Internet (IPv4) addresses, the UCSD telescope should see about 1/256 of those probes. We collect traffic data from the telescope using standard network interface cards in a PC, and CAIDA's Coralreef  software suite, which stores files in pcap format. As of December 2009, the network telescope captures in the range of 2GB up to and exceeding 100GB of compressed trace data per day. At the head-end of the capture process, the network's border router separates the legitimate traffic arriving at the telescope network (typically less than 1% of the total traffic volume) and forwards only non-legitimate traffic for monitoring and storage2. We have recently deployed a single host dedicated to storing approximately 30 days of data (at current traffic rates) from the UCSD Network Telescope. During normal operation the dataset covers a 30-most-recent-days moving window. Each day is represented by 24 compressed pcap files each containing one hour of data. Every hour the system automatically adds the most recent trace file to the collection, creating an almost real-time dataset with an effective latency of one hour. We periodically delete the oldest data to maintain the window without exhausting disk space. We hope to eventually increase our storage resources to maintain a larger window, since some event may not be recognized in its first 30 days of activity.
2.2 Reporting softwareWe provide a near-realtime, interactive graphical web interface  that displays rudimentary statistics of the telescope traffic. This coarse-grained view allows researchers to look for intervals of interest for further investigation. Figures 1 (a,b) show the cross-section of packets and bits by application (determined by TCP/UDP port number, hence a very rough estimate) for a week in mid-December 2009. The vast majority of packets and bytes are TCP SYNs to port 445 (which strongly dominates the NETBIOS port category). However, larger UDP packets (in blue) constitute an increasing fraction of recent traffic, observable in Figure 2a, which plots traffic volume (bit) over the last two years. Figure 2a also reflects the fact that in April 2009 we removed a 2Mbps rate-limiting filter on packets sent to certain targeted ports (in place to reduce measurement load, a method also suggested in ), to obtain more accurate estimates of the Conficker spread. Rather than re-instantiate the filter, we are transitioning to a real-time sharing model that will allow researchers to create their own filters on traffic. The right plot on Figure 2b shows the tuples (5-tuple flows ) per second colored by source country (geolocated with NetAcuity ) over the last 2 years, showing an increase in probes from China and Russia, as well as from IP addresses we could not geolocate to a country.
2.3 Prior data releasesOver the years we released a number of general telescope datasets such as Backscatter data  as well as curated datasets focused on specific security events: Witty Worm public and restricted data , and Code-Red Worms data . CAIDA has ameliorated the privacy risk of releasing victim host IP addresses and unexpected but occasional payload content with strict disclosure controls at some cost to research utility: (1) we deleted or sanitized payload; (2) we anonymized IP addresses of hosts using a common prefix-preserving technique. Our Acceptable Use Policies control re-identification risk by requiring that researchers agree to make no attempts to reverse engineer, decrypt, or otherwise identify the original IP addresses collected in the trace . Last year, we released two days of traffic traces from the network telescope in November 2008, as a baseline from dates prior to known Conficker activity, and then three days of data during Conficker growth [14,15]. These datasets represent the only raw network telescope data that we know of for use by the academic research community, and a sample of the kind of data we propose to make available in more useful form and timing to researchers.
2.4 Prior research and education contributions from telescope dataNetwork telescope data have allowed a few researchers - those with access to the data - visibility into a wide range of security-related events on the Internet, including misconfigurations malicious scanning of address space by hackers looking for vulnerable targets, backscatter from random source denial-of-service attacks, and the automated spread of malicious software called Internet worms [16,17,18,19,20,21]. In a 2004 study, Pang et al. termed unsolicited packets to unassigned Internet addresses "Internet Background Radiation ", and analyzed a broad sample of such traffic to find that, relative to legitimate traffic, it "is complex in structure, highly automated, frequently malicious, potentially adversarial, and mutates at a rapid pace." . Cooke et al. also described diversity in incoming traffic to ten unused address blocks  ranging in size from a /25 to a /8, announced from service provider networks, a large enterprise, and academic networks. Barford, et al.  found a bursty distribution of source addresses in darkspace, but consistency over time within a given chunk of dark space, and consistent (often vulnerable) destination ports probed, as expected for the targeted nature of most background radiation. Prior research contributions enabled by data from UCSD's network telescope have included studies of on DOS attacks [24,25], Internet worms  and their victims, e.g., Code-Red , Slammer , and Witty  worms. Data sets curated from telescope observations of these events became a foundation for modeling the top speed of flash worms , the "worst-case scenario" economic damages from such a worm , and the pathways of their spread and potential means of defense . CAIDA and other researchers continue studying the variety of traffic, and sources sending it, to empty address space. CAIDA also participates in DHS's Protected REpository for the Defense of Infrastructure against Cyber Threats (PREDICT) project, intended to promote empirical research into network infrastructure security. Partial support is provided by PREDICT for annotating and indexing telescope data in the PREDICT meta-data repository, which will enable us to leverage NSF infrastructure funds and add synergy and cross-agency momentum to this proposed project.
3 Proposed WorkOne of our most important lessons from Conficker is that in order to make the network telescope a more valuable resource for the next unpredictable large-scale outbreak of malicious network activity, we should be able to provide data access to vetted people with minimal curation and clean-up effort on our side, drastically reducing the lag time between data collection and analysis by researchers. In pursuit of this goal, we propose to enable experimental real-time data-sharing of telescope data. Our proposed tasks are: (1) enhance telescope data collection software and instrumentation to support automated generation of statistics, event triggers, and improved reporting focused discovery of emerging security threats; (2) deploy and evaluate a platform for real-time sharing of darkspace traffic data with vetted network and security researchers; (3) test a novel data-sharing framework and organize a workshop to get feedback from telescope data users.
3.1 Task 1: Enhance tools for telescope data analysis and visualizationTo increase the telescope usefulness as a way of detecting new security threats, we will extend our methodologies to support early detection and visualization of changes in the character of Internet background radiation. Specifically, we want to compute and report statistics at three layers: network, transport, and application. Network layer statistics include characterizing the geographic and topological origin of darkspace traffic. Transport layer (TCP and UDP header) information allows classification of behavioral patterns exhibited by individual IP source addresses. We have experimented with application-layer information extracted from the small fraction of darkspace traffic with payload to pattern-match against known set of application signatures, an area of stunted research due to the limited availability of traffic data. We expect that these tools will evolve based on research and experimentation enabled by our instrumentation enhancements.
Figure 3: Hilbert heatmap and geographic heatmap of IP addresses transmitting to telescope.
3.1.1 Network-layer analysisWe propose to enhance the network-layer analysis operationally provided by the telescope, focusing on the topological and geographic distribution of traffic sources as indicated by the source address in the IP header. Figure 3a presents an example of aggregated day of data in a graphical heatmap, mapping IP source addresses to IANA IPv4 allocation blocks using Duane Wessels' heatmap plotting software . The software maps the 1-dimensional IPv4 address space into a 2-dimensional image using a 12th-order Hilbert curve, so that CIDR network blocks appear as contiguous squares or rectangles in the image. Each point represents a single /24 network containing up to 256 hosts, and the color reflects observation of traffic from addresses in that /24. These heatmaps  insightfully illustrate the non-uniformity of source addresses of traffic to the UCSD telescope; over time (e.g., animations of) such maps can reveal anomalous events. Most observed source addresses belong to blocks allocated by the Regional Internet Registries: RIPE, APNIC, ARIN, LACNIC, and AfriNIC; fewer are from areas labeled `Various Registries', indicating allocations of pre-CIDR class B and class C IP address blocks before the mid-1990s. Even rarer are source addresses in the upper left of Figure 3a representing pre-CIDR class A blocks assigned to single organizations in the 1970s and 1980s. Almost no sources appear in Unallocated blocks, which suggests that (truly) randomly spoofed source addresses are an insignificant component of incoming traffic, consistent with previous work on darkspace source address distributions . The heatmap technique  has become popular in the security community for visualizing penetration of a given vulnerability or attribute into the IPv4 address space. We will incorporate heatmaps into the daily telescope monitoring and reporting software. Figure 3b maps the same data in geographic space on a world map.3 The source addresses are well-distributed across the globe, with approximate correlation to estimates of Internet users and usage in each country, confirming that unsolicited traffic remains truly a global phenomenon. Geographic reporting is particularly important for identifying the origins of cybersecurity threats.
3.1.2 Transport-layer analysis
|Group||Source Type||Description||% S||% p||kS||Mp|
|A||TCP||1||TCP port probe||TCP, many addrs, same port||14.66||53.08||242.59||64.67|
|A||TCP||2||TCP only > 1 port/addr||TCP, many addrs, > 1 port||1.39||2.02||23.03||2.46|
|A||TCP||3||TCP only 1 port/addr||TCP, 1 port for each addr||0.21||0.47||3.43||0.57|
|B||UDP||4||UDP port probe||UDP, many addrs, same port||27.31||1.74||451.84||2.13|
|B||UDP||5||UDP only, > 1 port/addr||UDP, many addrs, > 1 port||7.02||7.7||116.11||9.39|
|B||UDP||6||UDP only, 1 port/addr||UDP, 1 port for each addr||11.18||9.69||184.87||11.81|
|C||TCP+UDP||7||Both TCP and UDP||Mixed TCP and UDP||18.21||22.17||301.26||27.01|
|D||Conficker p2p||10||Only Conficker p2p||all pkts match Conficker p2p||3.53||1.30||58.39||1.58|
|D||Conficker p2p||11||Mixed Conficker p2p||only some Conficker p2p pkts||0.17||0.26||2.78||0.32|
|E||Other||8||Other Protocols||no TCP or UDP (mainly ICMP)||0.06||0.38||1.07||0.47|
|E||Other||9||Backscatter||TCP with ACK flag set||0.01||0.11||0.17||0.14|
|F||Unclassified||0||unclassified||source sent less than 20 pkts||16.24||1.08||268.71||1.31|
Figure 4: Hourly distributions of source lifetimes, packet rates and packet counts for four representative source types observed at the UCSD Network Telescope during the week Sat 3-Fri 9 April 2010 (UTC). Each type (1,4,7 and 10 in Table 1) shows notably distinct behavior patterns; although all have similar diurnal variations, their peaks appear at different values. Many sources are low-rate (below 0.01 p/s), but long-lived (lifetime peak near an hour), suggesting that they persist for periods greater than a single hour. TCP Probe sources are either long-lived low-rate or short-lived high-rate; UDP Probe sources typically send about 10 packets in 5 seconds; TCP+UDP sources resemble the longer-lived TCP Probe sources; and Conficker p2p sources generated mostly long-lived, low intensity flows.
3.1.3 Application-layer analysisGiven the privacy issues on host-populated networks, the telescope offers a unique opportunity to pursue payload-based traffic analysis, including evaluation and validation of traffic classification and modeling techniques. With respect for the sensitivities that remain in telescope data, we plan to carefully integrate application-layer analysis into our monitoring instrumentation.
3.2 Task 2: Deploy and refine hardware and software for real-time sharing of telescope dataOur previous model of indefinite storage and sharing of static, aged trace data on CAIDA servers [11,12,13] proved of some utility to cybersecurity researchers. Yet lessons learned from trying to share data during the Conficker attack motivate us to transition to a model of real-time data sharing with vetted researchers, storing a 30-day window (possibly up to 60 days since we recently increased our storage capabilities) of history. This new model should allow researchers timely access to a telescope observatory during a worm outbreak, where raw traces contain target addresses and payload that could enable quick autopsy of the structure and function of cybersecurity threats.
3.2.1 HardwareTo experiment with "bringing the code to the data" approaches, we will deploy a dedicated powerful data server capable to support several user accounts and their analysis tasks. We will also attempt to accommodate researchers who wish to provide their own computational resources, if our resources are insufficient for their needs. We propose to buy two fast network Juniper EX3200-48T-DC switches with two 10 gigE pluggable optics modules for each. These switches would replace our aging network hardware and raise our communication capacities to a new level thus making large volumes of data quickly and conveniently accessible for remote users and supporting additional equipment from other researchers.
3.2.2 User supportWe will enable infrastructural and administrative support to open and maintain access for vetted researchers to the telescope data server. Our goal is to run analysis programs on data within approximately an hour of its collection. To support long-term historical analyses of trends and creation of educational data kits, we will archive periodic hourly and daily traces similar to our current telescope operations, with specific intervals based on community feedback. We will invite researchers and security experts working with similar and related data to help us evaluate our data collection parameters and curation methods. The PS2 framework described in Task 3 will provide guidelines on what information researchers may reveal vs. anonymize in analysis results. We also anticipate that hands-on experience with real-time data analysis will prompt researchers to contribute to monitoring, analysis, and data visualization improvements to the telescope instrumentation (see also Section 3.3.2 on community outreach). We will evaluate such contributions to ensure that they meet appropriate ethics and safe data-sharing guidelines before integrating them into telescope operations.
3.3 Task 3: Community Development
3.3.1 Data-sharing policy frameworkWe have developed a Privacy-Sensitive Sharing Framework (PS2)  intended to effectively manage privacy risks that impede substantial data exchanges. The PS2 is a hybrid approach: a policy framework that applies proven and standard privacy principles between data seekers and data providers, coordinated with technologies that implement and enforce those obligations. The telescope data will be collected and made available as raw (unanonymized) traces, with payload (content). Rather than reduce the risk of releasing victim IPAs by anonymization and wholesale deletion of security-relevant data that also reduce the utility of the data to researchers, we will loosen the technical disclosure controls, and tighten use and disclosure obligations in the AUP. We will enforce purpose specification by obtaining explicit web form acknowledgment from the researcher that s/he will use the dataset solely for the stated research purposes. Researchers also will have to consent to use appropriate and reasonable care in safeguarding access to and preventing unauthorized use of the data, and are explicitly not permitted to share or transfer their access to the data. They are also prohibited from attempting to connect to, probe, or in any other way interacting or intervening with a machine or machine administrator identified in the dataset, without permission from CAIDA. For any publication or other disclosure, the researcher is obligated to anonymize or aggregate IPAs, network names, and domain names unless obtaining written authorization from CAIDA to do otherwise. We will guard access to the dataset(s) and authorization to use it by application review, approval and communication of acquisition instructions by CAIDA administrators. We will also require the researcher to report to CAIDA summary of the research and any findings, publications, or URLs using the data.
3.3.2 Workshop on telescope measurements and dataIn the beginning of Year 2 of the project we will host a workshop for researchers interested in using our telescope data, improving telescopes as scientific instrumentation, sharing their own telescope data, or sharing other types of data in real-time. This event, another in our series of popular Internet Statistics Measurements, and Analysis (ISMA) workshops , will help the community understand the breadth of research enabled by darkspace traffic data, as well as educate us regarding how much and what type of interest researchers have in darkspace traffic data, so we can maximize its utility for the research community. We will discuss lessons learned in this data-sharing experiment and publish a report summarizing findings and recommendations for use by other data providers and researchers.
4 Research and education opportunities enabledNetwork telescopes are one of the only types of Internet measurement instrumentation that allows world-wide global view of Internet security-related phenomena, providing opportunities for real-time and historical analyses otherwise inaccessible, and at relatively little cost. The proposed infrastructure will enable novel studies of Internet security threats crucial for understanding their provenance and developing efficient mitigating strategies. Our enhancements allow for research into network traffic in directions currently stunted for lack of access to such a data resource , including pursuit of the following questions:
- What type of traffic is hitting the telescope right now?
- Is aberrant traffic uniformly distributed across the address range?
- Can we reliably taxonomize unsolicited traffic into specific categories (e.g., misconfiguration, backscatter, scan, worm, attacks), extract statistically significant trends, and catch emerging behavior patterns?
- Which traffic characterization techniques will enable real-time anomaly detection?
- How does this telescope darkspace compare to other darkspace, in traffic volumes, characteristics, stability, and patterns?
- Can the proposed data-sharing model be used to allow such comparisons in real-time?
- Is it possible to detect the onset of security-related events by real-time correlation of distributed information from a variety of other security measurement instrumentation, such as intrusion detection systems which face local threats and honeynets which directly engage with malware? Can we at least semi-automate the generation of local filtering policies, by inferring the local vs global nature of events?
- How can we formally characterize the global nature and long-term trends in malware, e.g. worms, and the effectiveness of mitigation strategies, e.g. automated patching and malware removal? Do previous models of worm behavior match the unsolicited traffic observed by telescopes today, e.g., in terms of event duration and characteristics?
- Can the telescope provide ongoing monitoring of the worldwide usage and trends in cyberattacks? e.g., Are DOS attacks increasing in frequency and/or severity?
- Which lessons of telescope data sharing can extend to other datasets, e.g. remote processing / real-time access?
4.1 Why CAIDA is the most appropriate team for this projectCAIDA is a world leader in Internet measurement and data analysis of performance, workload, routing, topology, and security data , with years of experience in development, implementation, and evaluation of measurement infrastructure. We have been operating the UCSD telescope since 2001 and are committed to maintain and enhance this security research infrastructure. Support requested for this project combined with available DHS support for identifying, indexing, and annotating the most interesting subsets of data will ensure active life of the UCSD Network telescope until at least the end of 2013. SDSC provides access to advanced archival storage systems, data-handling platforms, and high-bandwidth networking. This institutional infrastructure will extend the lifetime of procured data collection even longer, preserving them as a useful resource for the security research community.
5 Broader Impact Activities
5.1 Dissemination of resultsThe results of this project will be broadly disseminated and presented to both academic and operational security research communities. We will advertise the availability of telescope data via conferences, meetings, web pages and CAIDA's blog, and by organizing the proposed workshop. CAIDA is also an active participant in DHS's PREDICT project, which aims to publicize and provide datasets for cybersecurity research. We will also collaborate with security experts seeking to merge diverse datasets into a comprehensive multi-faceted characterization of existing and emerging Internet security threats, in support of the global fight against malware.
5.2 Integration of research and educationSome of the most valuable training for future researchers is not found in carefully controlled classroom experiences - real world data has unexpected problems. Telescope data kits will enable invaluable hands-on experience in operationally relevant network security and traffic analysis research. PI Kc Claffy will use an ongoing stream of current telescope data as educational aids for her teaching, both in and out of the classroom. She will mentor a graduate student to be hired for this project, teaching him/her the basic principles of Internet measurements and providing practical insights into the difficulties in working with of massive (and messy) real Internet datasets. Day-to-day experience with collecting, processing and documenting research data will expose the student to the range of problems that security datasets can have and instill in him/her a high level of scrutiny and healthy skepticism of unusual results. CAIDA also has a successful track record of REU participation: in the last five years we supported and trained 17 undergraduates working on NSF-sponsored projects. We will apply for REU funds for this project as well. Dr. Claffy will collaborate with professors from the UCSD Department of Computer Science and Engineering, Prof. Brownlee (University of Auckland, New Zealand), and with post-doc Alberto Dainotti of University of Napoli Federico II (and sometimes visiting scholar at UCSD/CAIDA) to develop hands-on class projects using anonymized telescope data kits.
5.3 Supporting Diversity with CAIDA ActivitiesBased at UC, San Diego, CAIDA has a strong record of integrating diversity into our research activities. Since 1999, the composition of our 90 paid interns has included 25 females, 21 Asians and 4 Hispanics. We attract a diversified pool of graduate students. In addition to advertising the available position on the CAIDA web site and communicating with our connections among faculty members of the UCSD Computer Science and Engineering Department, CAIDA will post through various specialized UCSD sites such as UCSD Society of Women Engineers (ucsdswe.org), UCSD Student Chapter of the Society of Hispanic Professional Engineers (shpe.ucsd.edu). More importantly, resources like the UCSD Network Telescope and proposed real-time access mode are critical to the success of underrepresented groups in computer science and engineering, including women and minorities in other institutions. First, our telescope data are unique and the pool of unallocated IP addresses is rapidly shrinking, making it unfeasible to organize a similar collection of blackhole space data elsewhere. Second, the necessary infrastructure for data collection, curation, and storage is extensive and expensive, difficult to set up and requires large initial investments. Finally, acquisition of data typically involves personal trust relationships and people-networking with engineering and management personnel, which can leave underrepresented groups at a social disadvantage. Yet our project offers easy universal remote access to available data sets, thus contributing a leveling influence to the research and education playing field and facilitating the entrance of women and minorities into network security research.
- R. Pang, V. Yegneswaran, P. Barford, V. Paxson, and L. Peterson, "Characteristics of Internet Background Radiation," in Internet Measurement Conference (IMC), 2004.
- P. Porras, H. Saidi, and V. Yegneswaran, "Conficker," in SRI International Technical Report, 19 Mar 2000. http://mtc.sri.com/Conficker.
- P. Porras, H. Saidi, and V. Yegneswaran, "Conficker C P2P Protocol and Implementation," in SRI International Technical Report, 21 Sep 2009. http://mtc.sri.com/Conficker/P2P/.
- A. G. Z. Li, Y. Chen, and A. Kuzmanovic, "Measurement and Diagnosis of Address Misconfigured P2P Traffic," in INFOCOM '10, (San Diego, CA, USA), 2010.
- "UCSD Network Telescope," 2010. https://www.caida.org/data/passive/network_telescope.xml.
- D. Moore and K. Keys, "CoralReef software package." https://www.caida.org/tools/measurement/coralreef/.
- E. Kenneally and K. Claffy, "Dialing Privacy and Utility: A Proposed Data-sharing Framework to Advance Internet Research," IEEE Security and Privacy (S&P), July 2010. https://www.caida.org/publications/papers/2009/dialing_privacy_utility/.
- "UCSD Network Telescope Monitor," 2010. https://www.caida.org/data/realtime/telescope/.
- K. Keys, D. Moore, R. Koga, E. Lagache, M. Tesch, and k. claffy, "The architecture of CoralReef: an Internet traffic monitoring software suite," in PAM, 2001. https://www.caida.org/publications/papers/2001/CoralArch/.
- Digital Envoy, "Netacuity." http://www.digital-element.net/ip_intelligence/ip_intelligence.html.
- CAIDA, "Backscatter 2008 data." https://www.caida.org/data/passive/backscatter_2008_dataset.xml.
- CAIDA, "Witty data." https://www.caida.org/data/passive/witty_worm_dataset.xml.
- CAIDA, "Code red data." https://www.caida.org/data/passive/codered_worms_dataset.xml.
- "UCSD Network Telescope - Two Days in November 2008 Dataset," June 2009. https://www.caida.org/data/passive/telescope-2days-2008_dataset.xml.
- "UCSD Network Telescope - Three Days of Conficker Dataset," September 2009. https://www.caida.org/data/passive/telescope-3days-conficker_dataset.xml.
- M. Bailey, F. Jahanian, G. R. Malan, J. Nazario, D. Song, and R. Stone, "Measuring, Characterizing, and Tracking Internet Threat Dynamics," in OpenSig 2003, October.
- M. Bailey, E. Cooke, D. Watson, F. Jahanian, and J. Nazario, "The Blaster Worm: Then and Now," IEEE Security & Privacy, vol. 3, no. 4, pp. 26-31, 2005.
- D. Plonka, P. Barford, and V. Yegneswaran, "Internet Sink Deployments: On the Design and Use of Internet Sinks for Network Abuse Monitoring," in IEPG Meeting, July 2003. http://www.iepg.org/july2003/isink.pdf.
- T. Cymru, "Darknet project." http://www.team-cymru.org/Services/darknets.html.
- F. Jahanian, "Topology-Aware Internet Threat Detection Using Pervasive Darknets." http://www.eecs.umich.edu/fjgroup/topology/.
- A. Kumar, V. Paxson, and N. Weaver, "Exploiting underlying structure for detailed reconstruction of an Internet-scale event: Details of Witty worm," in Internet Measurement Conference (IMC), pp. 351-364, 2005.
- E. Cooke, M. Bailey, Z. Mao, D. Watson, F. Jahanian, and D. McPherson, "Toward understanding distributed blackhole placement," in Proc. ACM workshop on Rapid malcode, WORM '04, (Washington DC, USA), pp. 54-64, 2004.
- P. Barford, R. Nowak, R. Willett, and V. Yegneswaran, "Toward a Model for Source Address of Internet Background Radiation," in Proc. Passive and Active Measurement Conference, PAM '06, (Adelaide, Australia), 2006.
- D. Moore, G. M. Voelker, and S. Savage, "Inferring Internet Denial-of-Service Activity," Usenix Security Symposium, 2001.
- D. Moore, C. Shannon, D. Brown, G. M. Voelker, and S. Savage, "Inferring Internet Denial-of-Service Activity," ACM Transactions on Computer Systems, 2004.
- Alberto Dainotti, Antonio Pescapé and G. Ventre, "Worm Traffic Analysis and Characterization," in IEEE International Conference on Communications, June 2007.
- D. Moore, C. Shannon, and J. Brown, "Code-Red: a case study on the spread and victims of an Internet worm," in ACM Internet Measurement Workshop 2002, Nov 2002.
- D. Moore, V. Paxson, S. Savage, C. Shannon, S. Staniford, and N. Weaver, "Inside the Slammer Worm," IEEE Security and Privacy, vol. 1, no. 4, pp. 33-39, 2003.
- D. Moore and C. Shannon, "The Spread of the Witty Worm," IEEE Security and Privacy, vol. 2, no. 4, pp. 46-50, 2005.
- S. Staniford, D. Moore, V. Paxson, and N. Weaver, "The top speed of flash worms," in ACM Workshop on Rapid Malcode (WORM), pp. 33-42, 2004.
- N. Weaver and V. Paxson, "A worst-case worm," Workshop on Economics and Information Security (WEIS), June 2004. http://dtc.umn.edu/weis2004/weaver.pdf.
- D. Moore, C. Shannon, G. Voelker, and S. Savage, "Internet Quarantine: Requirements for Containing Self-Propagating Code," in INFOCOM03, 2003. https://www.caida.org/publications/papers/2003/quarantine/.
- D. Wessels, "IPv4-heatmaps." http://maps.measurement-factory.com/.
- A. Este, "IPv4 Address Space Maps of UCSD telescope data." https://www.caida.org/research/traffic-analysis/telescope-heatmaps/.
- E. Aben, "Conficker/Conflicker/Downadup as seen from UCSD Network Telescope, Feb 2009." https://www.caida.org/research/security/ms08-067/conficker.xml.
- M. Eisen and e. a. Michiel de Hoon, "Clustering 3.0." http://bonsai.hgc.jp/~mdehoon/software/cluster/software.htm.
- B. Alexander, "Port 137 scan." http://www.sans.org/security-resources/idfaq/port_137.php.
- E. Romang, "Activities on 49153/UDP linkproof.proximity.advanced ," April 2010. blog entry, http://eromang.zataz.com/2010/04/28/suc007-activities-on-49153udp-linkproof-proximity-advanced/.
- "Server queries." http://www.gnu-darwin.org/www001/ports-1.5a-CURRENT/games/qstat/work/qstat-2.11/info/a2s.txt.
- Songjie Wei and J. Mirkovic, "Correcting Congestion-Based Error in Network Telescope¿s Observations of Worm Dynamics," in Internet Measurement Conference (IMC), 2008.
- "Internet Statistics and Metrics Analysis Workshops." https://www.caida.org/workshops/isma/.
- Alberto Dainotti, kc claffy, Antonio Pescapé, "Issues and Future Directions in Internet Traffic Classification." submitted to IEEE Network August 2010.
- CAIDA. https://www.caida.org/publications/papers/.
Footnotes:1Pang et al. described ways to filter traffic to reduce storage and processing costs, all of which come at some expense to potential research utility, see Section 2. 2 The legitimate traffic is also a potential research resource, the sharing of which our PS2 framework  (cf. Section 3.3.1) could support. 3To map IP addresses to geographic locations we used Digital Envoy's NetAcuity service .
File translated from TEX by TTH, version 3.72.
On 14 Jun 2011, 11:50.