> -----Original Message----- > From: Craig R. McClanahan [mailto:[EMAIL PROTECTED] > Sent: Friday, September 26, 2003 9:38 PM > To: Struts Users Mailing List > Subject: Re: Going from https on /member.do, back to http on > index.jsp? > > > Jayaraman Dorai wrote: > > > > > > >>-----Original Message----- > >>From: Craig R. McClanahan [mailto:[EMAIL PROTECTED] > >>Sent: Thursday, September 25, 2003 10:55 PM > >>To: Struts Users Mailing List > >>Subject: Re: Going from https on /member.do, back to http on > >>index.jsp? > >> > >> > >>Max Cooper wrote: > >> > >> > >> > >>>Creating an index.do is one option. If it is not marked as > >>> > >>> > >>secure, sslext > >> > >> > >>>will write an absolute URL back to http:// for it. Many > Struts users > >>>advocate that all requests should be served by Actions, even > >>> > >>> > >>if the action > >> > >> > >>>merely forwards to a JSP. > >>> > >>> > >>> > >>> > >>I actually have a more important concern about doing this > -- exposing > >>information to people who can listen in on the packets that > >>go back and > >>forth. > >> > >>Many people would like, for example, to have the login form > >>be secure so > >>that your password is not sent across the wire encrypted, > but then go > >>back to http for the rest of the session for the better > performance. > >>The seemingly ideal solution is to have the destination of > the login > >>submit require SSL, and then switch back. Using a > >><user-transport-guarantee> security constraint, you can > even let the > >>container worry about the http->https transition for you. > >>Or, you can > >>use things like SSLext that is nicely integrated into Struts. > >> > >>However, the servlet spec offers no help in switching back to http, > >>because it is not a recommended practice. Why? Consider the > >>fact that > >>you now trust subsequent transactions from the same user, and are > >>typically using the session that was established earlier > (whether the > >>session was originally created in http or https turns out not to > >>matter). But the session id is no longer encrypted (it's > either in a > >>cookie or in rewritten URLs), so it can be easily forged by > >>anyone with > >>a network snooper between you and the server. How do you > >>guarantee that > >>an after-login request on that session is not being sent by > >>someone who > >>is forging the session id because they snooped it? > >> > >>I'm a conservative on security issues -- if I have an app > >>that needs SSL > >>sometimes, I arrange things to never accept a non-SSL > request on the > >>same session again, once I've switched to SSL (easy to do > >>with a Filter, > >>for example). If you're concerned enough to protect the > >>password, you > >>should be concerned enough to pay the CPU overhead for SSL > >>the remainder > >>of the session. Otherwise, it's likely that encrypting the > >>login form > >>will just give you a false sense of security. > >> > >> > >> > >>>I think you can also use the secure attribute in the sslext > >>> > >>> > >>tags to indicate > >> > >> > >>>whether the target of the link or form should be accessed > >>> > >>> > >>securely. In this > >> > >> > >>>case, you would add secure="false" to the sslext:link tag > >>> > >>> > >>that goes back to > >> > >> > >>>/index.jsp. > >>> > >>>-Max > >>> > >>> > >>> > >>> > >>Craig > >> > >> > > > >Will it be useful for cases like where credit card > information are sent through SSL and hence are secured. They > might snoop the session, if it is switched back to http, but > still the credit card information is not. Similarly, snooper > might steal the sessionId, but the password is not stolen and > is secured. Is that not better than nothing? Am I missing something? > > > The credit card information is not going back and forth > across the wire > unencrypted, but your session identifier is (for servlet based apps, > it's normally in a cookie or encoded into the URLs). > > In at least some ecommerce apps I'm familiar with, the credit card > number *is* stored in an HttpSession (or something analogous for > non-servlet technologies), so that the customer can go back > and buy some > more stuff, and do an expedited checkout the second time. If I know > that your app works that way, I can snoop the wire until the > user goes > to https (which I can't read, but can see happening) and then back to > http (which I can see). Now, I've got the cleartext version of the > session id, assume that the user has gone through the > checkout, and can > submit false additional purchase transactions to your app > based on that > session id, without having to resubmit a credit card number.
Will adding a checkout page which uses https, displaying the order details and asking the user to see all his order details again and then checking out (through https), make it secure. Essentially, the user can search, add, update and remove the shopping cart in the http mode and having the checkout page(displaying all the orders and then submitting) alone on https not secure enough? If it is not secure enough, will asking for a credit card number again through https solve it? > > Protecting just the password, but not the remainder of the current > session, is *not* better than nothing -- IMHO it is > substantially worse > because it gives you a *false* sense of security. Protecting the password limits the damage only to the session. If the password where to be a open text, the snooper can come back in any time with the stolen password, place the order for himself. That way, is protecting the password still not better than nothing? Taking a middle ground, between no https and all https, with a careful checks at critical points seems to make the application secured. > > Craig > > --------------------------------------------------------------------- To unsubscribe, e-mail: [EMAIL PROTECTED] For additional commands, e-mail: [EMAIL PROTECTED]
