e-Commerce security: Attacks and preventive strategies - IX

Site development best practices

This section describes best practices you can implement to help secure your site.

Security policies and standards

There are many established policies and standards for avoiding security issues. However, they are not required by law. Some basic rules include:

• Never store a user's password in plain text or encrypted text on the system. Instead, use a one-way hashing algorithm to prevent password extraction.
• Employ external security consultants (ethical hackers) to analyze your system.
• Standards, such as the Federal Information Processing Standard (FIPS), describe guidelines for implementing features. For example, FIPS makes recommendations on password policies.
• Ensure that a sufficiently robust encryption algorithm, such as triple DES or AES, is used to encrypt all confidential information stored on the system.
• When developing third-party software for e-Commerce applications, use external auditors to verify that appropriate processes and techniques are being followed.
• Recently, there has been an effort to consolidate these best practices as the Common Criteria for IT Security Evaluation (CC). CC seems to be gaining attraction. It is directly applicable to the development of specific e-Commerce sites and to the development of third party software used as an infrastructure in e-Commerce sites.

Security best practices remain largely an art rather than a science, but there are some good guidelines and standards that all developers of e-Commerce software should follow.

Using cookies

One of the issues faced by Web site designers is maintaining a secure session with a client over subsequent requests. Because HTTP is stateless, unless some kind of session token is passed back and forth on every request, the server has no way to link together requests made by the same person. Cookies are a popular mechanism for this. An identifier for the user or session is stored in a cookie and read on every request. You can use cookies to store user preference information, such as language and currency. This simplifies Web page development because you do not have to be concerned about passing this information back to the server.

The primary use of cookies is to store authentication and session information, your information, and your preferences. A secondary and controversial usage of cookies is to track the activities of users.

Different types of cookies are:

• Temporary cookies: These cookies are valid only for the lifetime of your current session, and are deleted when you close your browser. These are usually the good type. They are mostly used to keep your session information.
• Permanent cookies: These are for a time period, specified by the site, on the shopper's computer. They recall your previous session information.
• Server-only cookies: These cookies are usually harmless, and are only used by the server that issued them.
• Third-party cookies: These are usually used for tracking purposes by a site other than the one you are visiting. Your browser or a P3P policy can filter these cookies.

If you do not want to store cookies, here are other alternatives:

• Send user ID/password on every request: This was popular 5-10 years ago, but now recognized as an insecure technique. The user ID/password flowing under non-SSL is susceptible to attacks. This alternative is not practical for a high volume site. Pages that run under SSL would slow down site performance.
• SSL client side authentication: This is the most secure, but it is cumbersome for shoppers to install on their browsers. You have to pay for a company to verify who you are and to issue a certificate. The popularity of this technique for client-side authentication has decreased in recent years. It remains very popular on server sites.
• URL rewriting: This is a popular alternative to cookies. Each HTTP link on the page is specially encoded, but it is expensive for the site to implement. It interferes with the performance of the site because the pages cannot be cached and reused for different users. This alternative is susceptible to attack if it is not used under SSL.

Cookies marked as secure (storing encrypted data and passing to the user only under SSL) remain the most popular method of providing a secure online experience.

Using threat models to prevent exploits

When architecting and developing a system, it is important to use threat models to identify all possible security threats on the server. Think of the server like your house. It has doors and windows to allow for entry and exit. These are the points that a burglar will attack. A threat model seeks to identify these points in the server and to develop possible attacks.

Threat models are particularly important when relying on a third party vendor for all or part of the site's infrastructure. This ensures that the suite of threat models is complete and up-to-date.

Figure 11. Threat models
 

Responding to security issues

An effective overall security strategy is to be prepared when vulnerabilities are detected. This also means ensuring that software vendors selected for all or part of the site's infrastructure have proactive and reactive policies for handling security issues.

In the case of WebSphere Commerce, we can quickly form a SWAT team with key developers, testers, and support personnel. This becomes the highest priority for all involved parties. An assessment is made immediately, usually within the first few hours, to determine the vulnerability of the merchant's sites. A workaround or permanent solution is developed for the affected sites within a day. Then a "flash" issued to all customers to notify them of the problem, the solution, and how to check if they have been exploited. For critical issues, no one leaves until there is a solution.

Using an online security checklist

Use this security checklist to protect yourself as a shopper:

• Whenever you logon, register, or enter private information, such as credit card data, ensure your browser is communicating with the server using SSL.
• Do not shop at a site when the browser does not recognize the server's SSL certificate. This check is done by your browser the first time your URL becomes HTTPS for the site. If the certificate is not recognized, then your browser presents a pop-up message to inform you.
• Use a password of at least 6 characters, and ensure that it contains some numeric and special characters (for example, c0113g3).
• Avoid reusing the same user ID and password at multiple Web sites.
• If you are authenticated (logged on) to a site, always logoff after you finish.
• Use a credit card for online purchases. Most credit card companies will help you with non-existent or damaged products.
• A bricks and mortar store with an online brand is most likely a legitimate site. However, the site may still have vulnerabilities.

e-Commerce security: Attacks and preventive strategies - VIII

Defenses

Despite the existence of hackers and crackers, e-Commerce remains a safe and secure activity. The resources available to large companies involved in e-Commerce are enormous. These companies will pursue every legal route to protect their customers. Figure 6 shows a high-level illustration of defenses available against attacks.

Figure 6. Attacks and their defenses
 

At the end of the day, your system is only as secure as the people who use it. Education is the best way to ensure that your customers take appropriate precautions:

• Install personal firewalls for the client machines.
• Store confidential information in encrypted form.
• Encrypt the stream using the Secure Socket Layer (SSL) protocol to protect information flowing between the client and the e-Commerce Web site.
• Use appropriate password policies, firewalls, and routine external security audits.
• Use threat model analysis, strict development policies, and external security audits to protect ISV software running the Web site.

Education

Your system is only as secure as the people who use it. If a shopper chooses a weak password, or does not keep their password confidential, then an attacker can pose as that user. This is significant if the compromised password belongs to an administrator of the system. In this case, there is likely physical security involved because the administrator client may not be exposed outside the firewall. Users need to use good judgement when giving out information, and be educated about possible phishing schemes and other social engineering attacks.

Personal firewalls

When connecting your computer to a network, it becomes vulnerable to attack. A personal firewall helps protect your computer by limiting the types of traffic initiated by and directed to your computer. The intruder can also scan the hard drive to detect any stored passwords.

Secure Socket Layer (SSL)

Secure Socket Layer (SSL) is a protocol that encrypts data between the shopper's computer and the site's server. When an SSL-protected page is requested, the browser identifies the server as a trusted entity and initiates a handshake to pass encryption key information back and forth. Now, on subsequent requests to the server, the information flowing back and forth is encrypted so that a hacker sniffing the network cannot read the contents.

The SSL certificate is issued to the server by a certificate authority authorized by the government. When a request is made from the shopper's browser to the site's server using https://..., the shopper's browser checks if this site has a certificate it can recognize. If the site is not recognized by a trusted certificate authority, then the browser issues a warning as shown in Figure 7.

Figure 7. Warning to user
 

As an end-user, you can determine if you are in SSL by checking your browser. For example, in Mozilla® Firefox, the secure icon is at the top in the URL entry field as shown in Figure 8.

Figure 8. Secure icon in Mozilla Firefox
 

In Microsoft® Internet Explorer, the secure icon is at the bottom right of the browser as shown in Figure 9.

Figure 9. Secure icon in Microsoft Internet
 

Server firewalls

A firewall is like the moat surrounding a castle. It ensures that requests can only enter the system from specified ports, and in some cases, ensures that all accesses are only from certain physical machines.

A common technique is to setup a demilitarized zone (DMZ) using two firewalls. The outer firewall has ports open that allow ingoing and outgoing HTTP requests. This allows the client browser to communicate with the server. A second firewall sits behind the e-Commerce servers. This firewall is heavily fortified, and only requests from trusted servers on specific ports are allowed through. Both firewalls use intrusion detection software to detect any unauthorized access attempts.

Another common technique used in conjunction with a DMZ is a honey pot server. A honey pot is a resource (for example, a fake payment server) placed in the DMZ to fool the hacker into thinking he has penetrated the inner wall. These servers are closely monitored, and any access by an attacker is detected.

Figure 10. Firewalls and honey pots
 

Password policies

Ensure that password policies are enforced for shoppers and internal users. A sample password policy, defined as part of the Federal Information Processing Standard (FIPS), is shown in the table below.

Policy	Value
Account lockout threshold	6 attempts
Consecutive unsuccessful login delay	10 seconds
Matching user ID and password	N (no, they cannot match)
Maximum occurrence of consecutive characters	3 characters
Maximum instances of any character	4 instances
Maximum lifetime of passwords	180 days
Minimum number of alphabetic characters	1 alphabetic character
Minimum number of numeric characters	1 numeric character
Minimum length of password	6 characters
Reuse user's previous password	N (no, cannot be reused)

You may choose to have different policies for shoppers versus your internal users. For example, you may choose to lockout an administrator after 3 failed login attempts instead of 6. These password policies protect against attacks that attempt to guess the user's password. They ensure that passwords are sufficiently strong enough so that they cannot be easily guessed. The account lockout capability ensures that an automated scheme cannot make more than a few guesses before the account is locked.

Intrusion detection and audits of security logs

One of the cornerstones of an effective security strategy is to prevent attacks and to detect potential attackers. This helps understand the nature of the system's traffic, or as a starting point for litigation against the attackers.

Suppose that you have implemented a password policy, such as the FIPS policy described in the section above. If a shopper makes 6 failed logon attempts, then his account is locked out. In this scenario, the company sends an email to the customer, informing them that his account is locked. This event should also be logged in the system, either by sending an email to the administrator, writing the event to a security log, or both.

You should also log any attempted unauthorized access to the system. If a user logs on, and attempts to access resources that he is not entitled to see, or performs actions that he is not entitled to perform, then this indicates the account has been co-opted and should be locked out. Analysis of the security logs can detect patterns of suspicious behavior, allowing the administrator to take action.

In addition to security logs, use business auditing to monitor activities such as payment processing. You can monitor and review these logs to detect patterns of inappropriate interaction at the business process level.

The infrastructure for business auditing and security logging is complex, and most likely will come as part of any middleware platform selected to host your site. WebSphere Commerce, for example, has extensive capabilities in this area.

e-Commerce security: Attacks and preventive strategies - VII

Attacks

This section describes potential security attack methods from an attacker or hacker.

Tricking the shopper

Some of the easiest and most profitable attacks are based on tricking the shopper, also known as social engineering techniques. These attacks involve surveillance of the shopper's behavior, gathering information to use against the shopper. For example, a mother's maiden name is a common challenge question used by numerous sites. If one of these sites is tricked into giving away a password once the challenge question is provided, then not only has this site been compromised, but it is also likely that the shopper used the same logon ID and password on other sites.

A common scenario is that the attacker calls the shopper, pretending to be a representative from a site visited, and extracts information. The attacker then calls a customer service representative at the site, posing as the shopper and providing personal information. The attacker then asks for the password to be reset to a specific value.

Another common form of social engineering attacks are phishing schemes. Typo pirates play on the names of famous sites to collect authentication and registration information. For example, http://www.ibm.com/shop is registered by the attacker as www.ibn.com/shop. A shopper mistypes and enters the illegitimate site and provides confidential information. Alternatively, the attacker sends emails spoofed to look like they came from legitimate sites. The link inside the email maps to a rogue site that collects the information.

Snooping the shopper's computer

Millions of computers are added to the Internet every month. Most users' knowledge of security vulnerabilities of their systems is vague at best. Additionally, software and hardware vendors, in their quest to ensure that their products are easy to install, will ship products with security features disabled. In most cases, enabling security features requires a non-technical user to read manuals written for the technologist. The confused user does not attempt to enable the security features. This creates a treasure trove for attackers.

A popular technique for gaining entry into the shopper's system is to use a tool, such as SATAN, to perform port scans on a computer that detect entry points into the machine. Based on the opened ports found, the attacker can use various techniques to gain entry into the user's system. Upon entry, they scan your file system for personal information, such as passwords.

While software and hardware security solutions available protect the public's systems, they are not silver bullets. A user that purchases firewall software to protect his computer may find there are conflicts with other software on his system. To resolve the conflict, the user disables enough capabilities to render the firewall software useless.

Sniffing the network

In this scheme, the attacker monitors the data between the shopper's computer and the server. He collects data about the shopper or steals personal information, such as credit card numbers.

There are points in the network where this attack is more practical than others. If the attacker sits in the middle of the network, then within the scope of the Internet, this attack becomes impractical. A request from the client to the server computer is broken up into small pieces known as packets as it leaves the client's computer and is reconstructed at the server. The packets of a request is sent through different routes. The attacker cannot access all the packets of a request and cannot decipher what message was sent.

Take the example of a shopper in Toronto purchasing goods from a store in Los Angeles. Some packets for a request are routed through New York, where others are routed through Chicago. A more practical location for this attack is near the shopper's computer or the server. Wireless hubs make attacks on the shopper's computer network the better choice because most wireless hubs are shipped with security features disabled. This allows an attacker to easily scan unencrypted traffic from the user's computer.

Figure 4. Attacker sniffing the network between client and server
 

Guessing passwords

Another common attack is to guess a user's password. This style of attack is manual or automated. Manual attacks are laborious, and only successful if the attacker knows something about the shopper. For example, if the shopper uses their child's name as the password. Automated attacks have a higher likelihood of success, because the probability of guessing a user ID/password becomes more significant as the number of tries increases. Tools exist that use all the words in the dictionary to test user ID/password combinations, or that attack popular user ID/password combinations. The attacker can automate to go against multiple sites at one time.

Using denial of service attacks

The denial of service attack is one of the best examples of impacting site availability. It involves getting the server to perform a large number of mundane tasks, exceeding the capacity of the server to cope with any other task. For example, if everyone in a large meeting asks you your name all at once, and every time you answer, they ask you again. You have experienced a personal denial of service attack. To ask a computer its name, you use ping. You can use ping to build an effective DoS attack. The smart hacker gets the server to use more computational resources in processing the request than the adversary does in generating the request.

Distributed DoS is a type of attack used on popular sites, such as Yahoo!®. In this type of attack, the hacker infects computers on the Internet via a virus or other means. The infected computer becomes slaves to the hacker. The hacker controls them at a predetermined time to bombard the target server with useless, but intensive resource consuming requests. This attack not only causes the target site to experience problems, but also the entire Internet as the number of packets is routed via many different paths to the target.

Figure 5. Denial of service attacks
 

Using known server bugs

The attacker analyzes the site to find what types of software are used on the site. He then proceeds to find what patches were issued for the software. Additionally, he searches on how to exploit a system without the patch. He proceeds to try each of the exploits. The sophisticated attacker finds a weakness in a similar type of software, and tries to use that to exploit the system. This is a simple, but effective attack. With millions of servers online, what is the probability that a system administrator forgot to apply a patch?

Using server root exploits

Root exploits refer to techniques that gain super user access to the server. This is the most coveted type of exploit because the possibilities are limitless. When you attack a shopper or his computer, you can only affect one individual. With a root exploit, you gain control of the merchants and all the shoppers' information on the site. There are two main types of root exploits: buffer overflow attacks and executing scripts against a server.

In a buffer overflow attack, the hacker takes advantage of specific type of computer program bug that involves the allocation of storage during program execution. The technique involves tricking the server into execute code written by the attacker.

The other technique uses knowledge of scripts that are executed by the server. This is easily and freely found in the programming guides for the server. The attacker tries to construct scripts in the URL of his browser to retrieve information from his server. This technique is frequently used when the attacker is trying to retrieve data from the server's database.

e-Commerce security: Attacks and preventive strategies - VI

Points the attacker can target

As mentioned, the vulnerability of a system exists at the entry and exit points within the system. Figure 3 shows an e-Commerce system with several points that the attacker can target:

• Shopper
• Shopper' computer
• Network connection between shopper and Web site's server
• Web site's server
• Software vendor

Figure 3. Points the attacker can target 
 

These target points and their exploits are explored later in this article.