Alert Logic State of Cloud Security Report Fall 2012
An Empirical Analysis Of Real World Threats
CLOUD VS. ON-PREMISE SECURITY
Web Application Attacks Remain the Most Significant Threat for service provider environments.
Is the cloud inherently insecure? Are on-premise It environments a safer option for security-conscious organizations? Recognizing that these questions are critical to customers considering moving infrastructure into service provider-hosted and cloud environments, in February 2012 Alert Logic launched the first in a series of semi-annual reports on cloud security, based on analysis of threat data from its customers’ production environments. Reviewing 12 months of operational data, including more than two billion events and over 60,000 security incidents, Alert Logic concluded that the cloud is inherently no less secure than the on-premise environment.
For this second report, Alert Logic took the same approach: analyzing operational data from business customers in both on-premise and service provider environments and comparing the occurrence, frequency and diversity of incidents across seven categories of security threats.
The results of this study underscore Alert Logic’s earlier conclusion that the cloud is as safe as on-premise environments. Additional analysis shows that the type of infrastructure (service provider vs. on-premise) is a better determinant of the type and frequency of attacks than the target’s industry segment.
This second report also reinforces Alert Logic’s previous finding that Web application attacks are a significant threat for customers in all environments. Web application attacks are once again the number-one incident type experienced in the cloud, and the second most common incident type in on-premise environments. Based on these findings, we have taken a deeper look at Web application attacks in this report. Given that more than half of the Web application attacks we observed came from freely downloadable tools such as Havij, our findings reinforce the need to put a greater focus on this attack vector.
INSIGHT INTO ATTACKER BEHAVIOR
Targeted attacks are highly publicized. But our data suggests that the majority of malicious activity is opportunistic rather than targeted — attackers are looking for vulnerable targets, rather than selecting specific organizations to attack. This is demonstrated by:
• A high level of reconnaissance activity
• Geographic origins of reconnaissance activity
• The similarity of data across industries
The implication of these observations is that security managers should assume that vulnerabilities will be discovered and exploited, regardless of an organization’s likelihood of being specifically targeted.
KEY FINDINGS: INCIDENT OCCURRENCE, INCIDENT FREQUENCY, AND THREAT DIVERSITY
In this study period, service provider-managed environments did not encounter a greater level of threats than on-premise environments. All factors in the analysis supported this conclusion.
Web application attacks remain the most significant threat for service provider environments (53 percent of customers impacted) and the second most significant threat in on-premise environments (44 percent of customers impacted). Further analysis of Web application attacks reveals that the majority were perpetrated using common and freely available tools.
In the context of making precise comparisons between our first and second reports, it is worth noting that Alert Logic’s security research team makes ongoing improvements and refinements to its threat detection process. Because of this, there will always be changes between reporting periods that preclude direct quantitative comparisons with data analyzed in different periods. While direct quantitative comparison may not be possible, the two sets of data are similar, and differences should be looked at directionally, if not as precise measures.
The bottom line remains the same as in the initial report: While there are many factors to weigh when deciding whether to move infrastructure to the cloud, an assumption of insecurity should not be among them.
The first Alert Logic State of Cloud Security Report evaluated three factors—Incident Occurrence, Incident Frequency and Threat Diversity. We are continuing these three vectors of analysis in this report.
Incident Occurrence: The percentage of customers experiencing each type of incident. Customers are included if they experienced a specific class of incident at least once during the study period. (How many customers were impacted by each incident class?)
Incident Frequency: The average frequency of incidents of each type, per impacted customer. (How many times did impacted customers experience each type of incident?)
Threat Diversity: The threat diversity in each group, i.e., the number of unique incident types (of the seven classes reviewed) encountered by the customers in each environment. (How many different types of incidents did impacted customers experience?)
METHODOLOGY AND CUSTOMER DATA SET
Alert Logic Methodology
The data used in this report is actual incident data detected in customer environments secured by Alert Logic, not from surveys, lab environments or honeypots. Alert Logic captures security events in these environments through network-based, signature-driven intrusion detection systems (IDS). To correct for noise and false positives, Alert Logic utilizes a patented expert system that evaluates multiple factors in determining if one or more network-based events elevate to the level of an authentic security incident. Further, a team of GIAC-certified security analysts reviews each incident to ensure validity and to confirm the threat or compromise, providing an additional layer of scrutiny to minimize false positives.
As part of its ongoing process, Alert Logic regularly refines its threat detection based on changing threat activity. During this reporting period, Alert Logic fine-tuned the detection signatures used in its expert system, adding several new logic branches.
ON-PREMISE? HOSTED? SERVICE PROVIDER? CLOUD?
How Alert Logic categorized its customer data
In this report, Alert Logic studied incident data from over 1,600 customers in a broad range of industries. For its analysis, Alert Logic has categorized its data into two environments: on-premise and service provider. On-premise customers own and manage their own IT infrastructure. Service provider customers are an aggregation of all customers utilizing Infrastructure-as-a-Service solutions from a service provider, spanning from the elastic cloud to managed or dedicated hosted environments. On-premise deployments were typically larger than service provider deployments, featuring a broader set of applications and operating systems. The majority of both cohorts are located in North America and Western Europe.
GENERAL INSIGHTS BY ATTACK CLASS
Our research yielded the following general observations for the security incident categories used in our analysis:
DATA AND OBSERVATIONS BY INCIDENT CLASS
- Web application attacks and brute force attacks were the top two incident types for all environments.
- 1.5 billion security events observed during the study period were automatically evaluated and correlated through alert logic’s expert system and reviewed by alert logic’s security analysts.
- 70,000 security incidents were verified and classified into seven incident categories: application attack, brute force, malware/botnet activity, misconfiguration, reconnaissance, vulnerability scan and Web application attack.
Event Vs. Incident
Event: Evidence of suspicious behavior detected via an IDS signature.
Incident: An event or group of events that have been confirmed as a valid threat based on advanced automated analysis by Alert Logic’s expert system, and verified by certified analysts.
Web application and brute force attacks were the two most common incident types experienced in both on-premise and service provider environments. For service providers, reconnaissance was the third most significant attack vector. For on-premise environments, malware/botnet activity holds this position.
While the proportion of customers impacted in the top threat categories — Web application, brute force, reconnaissance and vulnerability scans — was not appreciably different between on-premise and service provider environments, we saw larger differences in other incident types. For malware/botnet activity, application attacks and misconfigurations, a far greater percentage of on-premise customers experienced incidents.
In each environment, certain types of incidents were more prevalent.
A higher percentage of customers in on-premise environments experienced brute force, vulnerability scan, application attack, malware/botnet and misconfiguration incidents.
For Web application attacks and reconnaissance incidents, a slightly greater percentage of service provider customers were affected. Service provider environments represent good targets for reconnaissance attacks, as scanning an IP range touches a large number of possible targets.
For every incident class, the number of incidents per impacted customer was higher in the on-premise environment. In some cases these differences were quite large.
While roughly half of all customers — whether service provider or on-premise — were likely to have experienced a Web application attack, the average number of such attacks was 61.4 among on-premise customers. For service provider customers, it was 27.8.
Brute force attacks, while similarly experienced by on-premise (46%) and service provider (39%) customers, also showed a wide disparity in frequency. On-premise customers averaged 71.7 such attacks, while impacted service provider customers were hit with an average of 42.6 brute force attacks.
For reconnaissance attacks, the difference was also significant. Service provider customers experiencing reconnaissance attacks averaged 2.7 such attacks; for on-premise customers, the average was 40.
For on-premise environments, customers experienced an average of 2.9 types of incidents. Service provider customers experienced an average of 2.1 incident types. While lower threat diversity does not necessarily indicate an inherently safer environment, it does suggest a narrower range of threats, requiring a different security posture.
In this second State of Cloud Security Report, Alert Logic added two new areas of analysis:
1. Industry Segmentation
2. Country of Attack Origin
Alert Logic categorized its customer data along industry lines, grouping customers into twelve categories. The incident occurrence for these industries, across both environments, is shown below. The top five industries on the list yielded enough data to make meaningful observations.
Alert Logic researchers made the following observations:
In general, the type of infrastructure (on-premise versus service provider) is a more important determinant of type and frequency of attack than the targeted customer’s industry sector.
Businesses with on-premise IT environments consistently experienced more frequent attacks across a more diverse set of threats, compared to businesses with cloud-based IT infrastructure.
The divergence in attack frequency between on-premise and service provider environments was wide:
- For IT Services, impacted service provider companies experienced 17.6
- Web application attack incidents per customer, compared to 159.2 for on-premise environments.
- For E-commerce, the gap was also significant: 29.5 for service providers versus 103.3 for on-premise. (However, the percent of customers impacted was almost twice as large in the service provider E-commerce sector than in the same sector for enterprises.)
- Other instances where the on-premise attack frequency far outweighed that among service provider customers were reconnaissance incidents in E-commerce: 129.5 (on-premise) versus 1.7 (service provider); brute force attacks in IT Services: 218.4 (on-premise) versus 25.8 (service provider); and reconnaissance attacks in Financial Services: 62.6 (on-premise) versus 1.6 (service provider).
While industry differences were less significant than the type of environment, we observed some interesting industry-specific variations:
IT Services companies that have a public presence experienced a large number of Web application attacks. There are also constant brute force attempts to gain access to these environments, likely because of the number of individuals with escalated privileges, access to servers and network configurations that make them useful to attackers.
E-commerce and SaaS environments are frequent targets of scanning, much of which is due to compliance assurance by third-party testers.
Malware is seen more frequently in Financial Services, possibly due to large numbers of desktop endpoints in banking environments and the volume of financial transactions conducted daily from personal computers.
A significant number of Media companies are being targeted by hacktivists with a public agenda against government regulation of the Internet.
Energy and Healthcare enterprises seem to be among the most heavily targeted industry segments, but Alert Logic researchers consider this conclusion to be preliminary based on the lower representation of these industries in the data set.
During this report period, Alert Logic recorded incidents originating in 165 countries, with the U.S. and China accounting for nearly half of the total number. The top ten countries were:
The United States was the country of origin for 33 percent of the incidents analyzed in this study and is responsible for 35.4 Web application attacks per impacted customer. We believe that this may be the result of the U.S. having a large number of unsecured personal computers with access to broadband connections that are attractive targets for hackers in other countries as well as botnet attacks.
China accounted for 16 percent of the attacks, ranking second. We noted an especially high frequency of incidents per customer impacted for reconnaissance attempts originating in China. This suggests a scenario in which hackers in China are doing reconnaissance, identifying vulnerable workstations in the bandwidth-rich U.S., adding those machines to botnets and using them to launch attacks on nearby targets.
China, India, The Republic of Korea and Russia were the source of many reconnaissance and brute force attacks, perhaps being launched from compromised machines in regions that tend to have older, unpatched operating systems.
Germany has one of the largest concentrations of command and control servers in the world. This is a result of a large number of hacking groups residing in an area, combined with the increasing number of regional data centers. Russia has a large number of application attacks driven by an active black market for binary exploit code. India and The Republic of Korea are significant contributors to incidents only because of brute force attempts.
Threat diversity by country is universally low: among customers attacked from one country, most experience fewer than two types of incidents from that country. This is lower than the threat diversity for both on-premise (2.9 different types of threats) and service provider (2.1 different threat types) customers.
There are also different types of and motivations for attacks in different regions; for example, reconnaissance to access data and proprietary information from China versus profit-driven application attacks from Russia.
TWO-THIRDS OF ALL ATTACKS CAME FROM FREELY DOWNLOADABLE TOOLS.
DRILLING DOWN ON WEB APPLICATION ATTACKS
Web application attacks are prevalent and frequent, and are implicated in many data breaches. In this report, we analyzed the tools used to launch these attacks and found that a majority of attacks are launched via identifiable tools, many of which are free, easily available and easy to use.
There are some differences between service provider and enterprise environments: 20 percent of service provider customers experience attacks that use Havij, compared to only 8 percent in enterprise environments. On the other hand, impacted enterprise customers typically experienced four times more incidents, compared to impacted service provider customers.
Some of the attack tools used are open source or easily obtained (Havij, SqlMap, NetSparker, Nessus, w3af, loginpwnz). Others are vulnerability-specific or software-specific attacks (shopxp scanner, tomcat scanner). Still others are less easily identified and more generic in nature. The popularity and ease of use of Havij sets it apart from the rest in terms of the threat is poses; it enables relatively unsophisticated attackers to launch attacks. SqlMap is one of the most powerful SQL injection tools, but requires a more skilled attacker to execute effectively.
The impact of a successful attack for each of these tools varies significantly. A successful Havij or SqlMap (or loginpwnz) attack, for example, results in the download of an entire database, and even the ability to run code on a SQL server. Conversely, a successful w3af, nessus or generic incident would only identify a vulnerability, and would not cause direct damage.
Occurrence: Source Of Web APP Attacks:
FROM RECON TO BREACH: GLOBAL ROADMAP OF A COMPROMISE
THE DATA TELLS THE STORY
In its initial State of Cloud Security Report (February 2012), Alert Logic found that differences in the types and frequency of incidents experienced in the on-premise and service provider environments did not align with general perceptions about cloud security. In fact, service provider environments showed lower occurrence rates than on-premise environments for every class of incident examined.
In this report, we find once again that the service provider environments where cloud infrastructure is hosted are not inherently less secure than on-premise environments.
With the addition of industry segmentation, Alert Logic finds that variations in threat activity among industries are less important than the environment where infrastructure is located. It is not safe to assume that one’s industry is not targeted by attackers, or that an organization is too small to be targeted. Alert Logic’s analysis suggests that attackers are using reconnaissance techniques to identify and exploit vulnerable targets wherever they can be found. Security planning should consider a range of technologies in order to provide defenses against these opportunistic threats. Alert Logic’s analysis also demonstrates that attacks originate from all over the world, and simple geographic filtering is not a reasonable countermeasure.
The most significant finding of the report is the prevalence of Web application attacks. With the application vulnerabilities remaining commonplace and the growing availability and ease of use of automated attack tools, we expect to see the number of these attacks continue to rise, as they no longer require an especially skilled attacker. This observation is underscored by a study cited by analyst Ramon Krikken at the recent Gartner Security & Risk Management Summit. He noted a report showing that nearly one-third of all Web applications are vulnerable to SQL injection attacks, and recommended considering use of defense tools such as Web application firewalls (WAFs), along with secure application development practices.
Organizations should make sure that they use secure application development processes and tools to identify application vulnerabilities. Further, as Krikken cautioned, given the high instance of cross-site scripting flaws, organizations need to go beyond relying on continuous development, testing and implementation of software patches—human interventions that are vulnerable to human error. Instead, they should consider active WAF technologies to block attacks. Access to security research and intelligence that informs security planning is also an important element of proactive defense.
Given the prevalence of unsophisticated attacks, such as brute force and reconnaissance, in both cloud and on-premise environments, and across all industries, the fundamentals apply: multi-layer security, close attention to basic management practices, such as patch management and upgraded operating systems, and use of monitoring and defensive technologies to identify and stop attacks.
When selecting cloud service providers, enterprises should consider the rigor and application of these fundamentals in their evaluation process. As we noted in the Spring 2012 State of Cloud Security Report, it is the quality of management applied to any IT environment that drives good security.
For the appendix with complete data tables, download the whitepaper.
Lead Researcher: Stephen Coty
Lead Analysts: Michael Bentley, Tyler Borland, Mukul Gupta, PhD, Charles Tarun
Editors: Maureen Rogers, John Whiteside
© 2014 Rackspace US, Inc.
Except where otherwise noted, content on this site is licensed under a Creative Commons Attribution-NonCommercial-NoDerivs 3.0 Unported License
See license specifics and DISCLAIMER