Implementer's notes

What might get caught in the gears under the hood?

Subscribe to RSS feed

Posts tagged with "Security"

Hvor sikker er egentlig norsk netthandel?

, , , ...

[To those of my readers who understand English, but not Norwegian: This article is written in Norwegian because it is about the use of encryption in Norwegian online shopping sites. I may post an English version later]

Handel på nett utgjør en stadig større del av norsk handel, og ble ifjor estimert til 48 milliarder kroner av DIBS, ifølge en artikkel på Digi.no/NTB.

Når så mye penger er i omløp er det en nødvendighet at man har tillit til sikkerheten i nettbutikkene.

Sikkerhet i nettbutikker

Sikkerheten som man må ha tillit til består av flere deler, hovedsaklig:

  • Hvor sikker oppleves selve handelen?
  • Hvor sikkert håndteres dine personlige data?
  • Hvor sikkert håndteres betalingen?

Disse delene avhenger delvis av hva kunden selv gjør, for eksempel ved valg av passord, deretter av hvordan nettbutikkens funksjonalitet mot kunden er lagt opp, for eksempel bruk av kryptering, og til sist av hvordan de interne systemene er sikret.

Kundens lokale sikring av egen informasjon er kundens eget ansvar. Selv om tekniske hjelpemidler som sikkerhetsprogramvare og passordkalkulator kan hjelpe, er det i hovedsak bare opplæring og brukerens generelle årvåkenhet som nytter på dette området.

Tilsvarende, nettbutikken er ansvarlig for sikring av interne systemer og den lagrede informasjonen. For utenforstående er det vanskelig, om ikke umulig, å vite om nettbutikken faktisk har sikret systemene sine skikkelig, og man må som kunde ta det for gitt at de er sikret godt nok. Dersom nettbutikken mot formodning ikke har sikret systemene skikkelig er vanligvis den første indikasjonen man som kunde får om at det har vært noe galt at det publiseres store artikler i avisene om datainnbrudd hos nettbutikken. Man kan imidlertid få hjelp til å vurdere hvor gode systemer nettbutikken har dersom nettbutikken har blitt sertifisert av en uavhengig tredjepart, f.eks. organisasjonen Trygg E-handel.

Det eneste området av sikkerheten hvor man som kunde rimelig enkelt kan avgjøre om sikkerheten i nettbutikken er god nok gjelder hvordan nettbutikken er bygget opp og fungerer når du handler der.

Bruk av kryptering i nettbutikker

Når man handler i en nettbutikk går man gjennom flere steg:

  1. Undersøke vareutvalget
  2. Velge varene man vil kjøpe
  3. Sjekke hva som er i handlevognen
  4. Registrere seg som kunde, eller
  5. Logge inn dersom man allerede er kunde
  6. Gå til kassen med handlevognen
  7. Oppgi hvordan varene skal leveres
  8. Betale for varene

Min personlige oppfatning er at alle disse aktivitetene bør gjøres over en kryptert (sikker HTTPS) forbindelse, men det er nesten ingen nettbutikker som har implementert dette. Derimot er det ganske viktig at aktivitet nummer 4 til og med nummer 8 alltid gjøres over en kryptert forbindelse for å hindre at sensitiv informasjon kommer på avveie.

Etter at jeg i desember 2011 kom over en nettbutikk som brøt flere fundamentale regler for god sikkerhet i krypterte nettsider har jeg sett nærmere på hvor sikre flere grupper av nettbutikker er når det gjelder de første 6 aktivitetene nevnt ovenfor, og i løpet januar har jeg gjort en slik enkel undersøkelse av 34 norske, og 12 britiske nettbutikker. I denne undersøkelsen har jeg ikke sett på hva som skjer når man oppgir hvordan varen skal leveres og hvordan man skal betale (aktivitet 7 og 8), siden dette krever at man er registrert som kunde, og at man gjennomfører hele kjøpsprosessen, noe jeg ikke var interessert i å gjøre som del av denne undersøkelsen.

Utvalget av nettbutikker var stort sett kjente merkenavn i flere forskjellige markeder, og de fleste har også vanlige utsalgssteder.

Hver butikk ble sjekket for hvert aktivitetspunkt, og hvorvidt den nettsiden man kom til, og alle ressurser lastet i denne siden var kryptert. Hvis enkelte ressurser, selv bilder, ble lastet uten kryptering ble hele siden kategorisert som "ukryptert", siden det kreves omstendelige undersøkelser fra brukerens side for å finne ut hvorfor siden er markert som usikker. Dette gjaldt også sider hvor den krypterte siden var "gjemt" inne i en ramme på en ukryptert side.

Resultatet av denne undersøkelsen er mildt sagt skuffende:

  • Bare to norske butikker, Boots og Kitchn, krypterte alle aktiviteter, mens ingen av de britiske gjorde dette.
  • Over halvparten (52%) av de norske nettbutikkene benyttet *ikke* kryptering i det hele tatt (eller ikke godt nok) for de aktivitetene som ble testet, tilsvarende for de britiske butikkene var 16%.
  • 26% av de norske butikkene krypterte aktiviteten for registrering, innlogging, og gå til kassen, men 67% av britiske butikkene gjorde dette. Inkluderer man også sikker visning av handlevognen ble tallene henholdsvis 12% og 8%.
  • 35% hadde sikker registrering som kunde, sammenlignet med 75% blant britiske butikker
  • 26% hadde sikker innlogging, sammenlignet med 75% blant britiske butikker
  • 35% hadde sikker start på å gå til kassen, sammenlignet med 75% blant britiske butikker. Gå til kassen innebærer som oftest at man må logge inn.
  • I en av de norske, og en av de britiske butikkene var disse tre aktivitetene riktignok kryptert, men inkluderte ukrypterte ressurser som åpnet for at en ondsinnet person kan ta full kontroll av den krypterte siden. Ytterligere to norske butikker blandet inn mindre farlig ukryptert innhold.
  • I flere av de nettbutikkene som ikke benyttet kryptering for en aktivitet, men hvor det var mulig å aktivisere kryptering, viste det seg at disse variantene lastet ukryptert innhold, og i noen tilfeller åpnet
    dette sårbarheter i den krypterte siden.

Hvorfor bruke kryptering?

Hvorfor skal aktiviteter som registrering som kunde, innlogging, og gå til kasse alltid foregå over en kryptert forbindelse?

Svaret er at disse aktivitetene i det minste omfatter bruk av brukernavnet og passordet ditt, og kan også omfatte andre private detaljer som adresse og telefonnummer, eller adgang til disse, og noen ganger kan nok informasjon om personnummer være involvert. I mange tilfeller lagrer også nettbutikken informasjon om kredittkortet ditt i profilen din, slik at du slipper å taste det inn hver gang du handler, og i så fall kan en ondsinnet person som får tak i passordet være i stand til å endre profilen din, og gjennomføre handler med ditt kredittkort.

Når noe sendes ukryptert vil hvem som helst som er i kontroll av en ruter mellom deg og nettstedet, for eksempel i et trådløst nettverk, være i stand til å avlytte det som sendes, men kan også enkelt endre innholdet i det som sendes, og for eksempel legge inn kode som sender kopi av det som tastes inn til angriperen.

En mulig måte et slikt angrep kan gjennomføres på, er at noen plasserer et "gratis WiFi" nettverk ved et trafikk-knutepunkt hvor mange setter seg ned med mobilen eller PCen sin. Med den rette programvaren installert i nettverket kan man enkelt avlytte eller modifisere all ukryptert trafikk gjennom nettet.

I tre av de norske butikkene og to av de britiske ble brukernavn og passord som ble oppgitt på ukrypterte innloggings-sider riktignok sendt kryptert, men dette gir bare et bedragersk skinn av "sikkerhet", siden en ondsinnet person bare trenger å endre innholdet i den ukrypterte siden slik at den sender en kopi av brukernavn og passord til ham selv, i tillegg til den faktiske innloggingen til butikken. Det er egentlig ganske utrolig at man i 2012 fremdeles kan se slike innloggingssider i bruk.

I en av de undersøkte butikkene ble riktignok kryptering benyttet for flere aktiviteter, men dette var gjemt inne i en ramme (frame) som var laget av en ukryptert side, og den såkalt sikre siden inne rammen, hvor man kan bli bedt om personnummeret når man har valgt enkelte betalingsalternativ, brøt også de fleste reglene for hvordan sikre sider skal lages og åpnet for enda flere sikkerhetsproblemer. Effektivt vil det si at aktiviteten foregikk over en ukryptert forbindelse siden det ikke er enkelt å finne ut at rammen bruker kryptering (eller har flere andre sårbarheter), og en ondsinnet person kan veldig enkelt endre hva som skal vises i rammen, siden informasjonen om det blir sendt ukryptert og kan endres uten at det kan oppdages på noen enkel måte. Dette er også et annet eksempel på noe som ikke burde bli brukt av en seriøs nettbutikk i 2012.

Det er imidlertid ikke bare bruken av kryptering som avgjør om forbindelsen til en nettbutikk er sikker. Som en del av undersøkelsen sjekket jeg også status for en velkjent, men mindre synlig sårbarhet i de sikre tjenerne for de nettbutikkene som benyttet slike. Jeg fant da ut at bare 29% av nettstedene deres har oppgradert slik at de er beskyttet mot det såkalte "Renego" angrepet , som ble oppdaget, og fikset for 2 år siden, til sammenligning har 67% av de britiske nettbutikkene fikset problemet; og generelt har ca. 65% av alle sikre nettsteder fikset dette problemet. 53% av de norske butikkene som ikke er oppgradert er dessverre også slik konfigurert at de er sårbare mot det fulle "Renego" angrepet, hvor angriperen kan lure nettstedet til å utføre oppgaver med kundens ID. Forøvrig, norske nettbanker har det samme problemet, kun en av de 6 jeg har sjekket, Skandiabanken, har fikset problemet, og de fleste andre er i tillegg sårbare for angrepet (En stor bank har delvis fikset problemet i løpet av de siste ukene, men ikke fullført dette ennå).

Hva kan gjøres for å forbedre situasjonen? Først og fremst må nettbutikkene fikse nettstedene sine, men for å få god sikkerhet i det lange løp må man få på plass gode kontrollrutiner, for eksempel i form av sertifiseringsordninger, og ved at kundene holder et øye med sin egen sikkerhet.

Sertifiseringsordninger

Som nevnt ovenfor er det normalt vanskelig for utenforstående å kontrollere hvor godt en nettbutikk ivaretar sikkerheten. Dette skyldes at man må ha adgang til informasjon som de færreste bedrifter offentliggjør, av forretningshensyn. Det er imidlertid mulig at en uavhengig tredjepart undersøker bedriften og dens oppfyllelse av nødvendige krav, og dersom disse er oppfylt utsteder et sertifikat om dette.

Etter min oppfatning må disse kravene inkludere et krav om at kryptering brukes for all funksjonalitet som håndterer sensitiv informasjon må være kryptert.

Organisasjonen Trygg E-handel, som er en norsk sertifiseringsordning, har blant sine krav til nettbutikker at "Personopplysningene skal beskyttes mot ekstern inntrenging", noe som etter min oppfatning betinger at all håndtering av slik informasjon må foregå kryptert, siden man ikke kan beskytte et system mot "ekstern inntrengning" dersom passordet er enkelt tilgjengelig i ukryptert form på nettet. Likevel, to av de fire nettbutikkene i min test som var sertifisert av Trygg E-handel benyttet ikke kryptering for innlogging og registrering.

Hva kan kundene gjøre for å sikre seg at nettbutikken bruker god kryptering?

Først og fremst må kundene gjøre seg kjent med hvordan nettleseren deres indikerer at en side er kryptert. Denne indikasjonen har varierende utseende og plassering i de forskjellige nettleserne, og utseendet kan også variere med hvilken type sertifikat (identifikasjonskort) nettstedet presenterer til nettleseren, men er vanligvis assosiert med en "hengelås", slik som i Opera og Internet Explorer, men ikke alltid, siden Firefox benytter et litt annet utseende. I tillegg vises en spesiell, grønn variant av "hengelåsen" når nettstedet bruker et såkalt EV (Extended Validation) sertifikat, hvor informasjonen om nettstedet har blitt kontrollert spesielt grundig; slike sertifikater brukes blant annet av nettbankene, og bør brukes av alle nettsteder som krever betaling for varer og tjenester.

Bildet under viser eksempler på hvordan et ukryptert nettsted, et kryptert, og et nettsted med et Extended Validation sertifikat blir vist i forskjellige nettlesere.



Når en kunden går til en nettbutikk må han eller hun sjekke "hengelåsen" før de starter inntasting av brukernavn og passord, eller annen personlig informasjon, for eksempel under registrering, og før man starter betalingsprosessen. Dersom "hengelåsen" ikke vises skal man avbryte det man holder på med siden informasjonen ikke blir håndtert sikkert nok. Man bør også informere nettbutikken om hva man har observert.

Oppsummering

Basert på det som er vist ovenfor, vil jeg påstå at både den manglende bruken av kryptering, og den manglede oppgraderingen av nettstedene for å fikse kjente sikkerhetsproblemer, gir en klar indikasjon på at norske nettbutikker dessverre ikke er sikre nok.

Det kan også stilles spørsmåltegn ved om sertifiseringsordningene passer godt nok på dette området.

Manglende bruk av kryptering for sensitiv informasjon sender et signal til kundene om at man ikke har tenkt godt nok igjennom sikkerheten for nettstedet. Dette signalet er forhåpentligvis feil, men etter min mening kan det skape berettiget tvil om hvor godt sikkerheten er ivaretatt internt i butikken. Eller for å si det på en annen måte: Det er bra å låse bakdøren, men det nytter lite dersom hovedinngangen er vidåpen.

Jeg vil på det sterkeste oppfordre norske nettbutikker om å ta sikkerhet og bruk av kryptering på alvor. Dersom nettstedet ikke bruker kryptering for sensitiv informasjon bør det fikses raskest mulig, og dersom det ikke kan gjøres umiddelbart bør strengt tatt nettstedet tas av luften inntil problemet er fikset.

Det er synd å si det, men dessverre tror jeg at det ikke blir noen stor bedring før noen gjennomfører et angrep, eller kundene blir mer observante, og stemmer ved å flytte sine museklikk til nettbutikker som passer bedre på kundenes sikkerhet.

Temporary workarounds are ... well ... temporary

, , , ...

When the TLS Renego vulnerability was discovered a year ago (November 4, 2009), many SSL/TLS server vendors rushed out a temporary workaround to mitigate the threat against servers: Disable renegotiation by default on the server.

This had the benefit of preventing the easiest way of using this new attack vector: Inserting a request into the server's command stream, before letting the client take over and receive the malicious result.

It did not, however, help clients if the attack was staged against them, but such an attack is more difficult to accomplish and does not look any different from an ordinary certificate replacement attack, except when the server requires client certificate authentication.

Since the real Renego protocol patch (the RI Extension, RFC 5746) was released in February 2010, and Opera 10.50 was released with this update, I have occasionally received complaints from users and system administrators about Opera's security information item "The server does not support secure TLS renegotiation", claiming that we should not display that for their server because they have added the above mentioned workaround. One of the references used for these was Ivan Ristic's SSLlabs tester. Ivan has since updated the site, and also posted an article about this topic

At least two server vendors have used the same argument about why they do not need to immediately ship an update supporting the RI extension.

I suspect that this perception, that the workaround is "sufficient", is delaying the deployment of updates with the RI-extension, so it is time to set the record straight:

Disabling server-side renegotiation was a quick & dirty, and very temporary, workaround deployed while there was no other, and more secure options available, in order to mitigate the discovered problem. It was never meant to be a permanent solution, nor does it provide any real security.

One reason for this is an aspect of the Renego-problem that many forget: The attack can be used against the client, too, not just the server! Admittedly, the client-side attack is more difficult to carry off, and will usually be indistinguishable from a normal Man-In-The-Middle attack with a fake certificate, but there might still be situations where such an attack can yield results for an attacker, even against a server that has disabled renegotiation, because the clients cannot disable that functionality.

But the other reason this is a significant problem is that the client cannot know that the server has implemented the workaround! It have to treat any server that does not return the RI-extension as if it is unsecure. Even if the client should waste time probing the server to "confirm" that the server refused to renegotiate, the result would be inconclusive, for two reasons:

  • An attacker can fake the response, particularly the aggressive "close the connection" response. So the client might think the server is "secure", while it isn't.
  • Some servers do not accept a client-initiated renegotiation, but many servers, particularly ones requiring client certificate authentication, will tell the client that it wants to renegotiate the connection. Such server-initiated renegotiation is usually triggered in response to specific queries to the server, and these server-specific triggers are generally unknowable to a client trying to perform a general capability test of the server. So, once again, a client might think that the server is "secure", while it isn't.


For these reasons, as well as the fact that such testing would waste time, no client have, to my knowledge, realistically considered probing the server. Doing so would be a waste of time and resources and would obtain a meaningless result.

Even worse, however, is that a recently released client that support the RI-extension cannot know whether the connection with an unpatched server has been intercepted and is being manipulated, because without the RI extension there is no way to tell securely that the client and server have only been talking to each other, and not also an additional party.

Therefore, all server and OS vendors that still haven't released a Renego-patch for all their maintained versions (beta versions do not count): It is time to get down from the fence and release a patch. Now!

Is it time to block unsecured content in "secure" documents?

, , , ...

Over the past few years browser vendors have been steadily raising the bar for what can be called secure connections and documents. SSL v2 has been phased out, as has 40- and 56-bit encryption. Extended Validation (EV) has been introduced to provide better identity assurance, and the W3C is about to release the Web Security Context UI specification to make the security indicators more consistent.

There is, however, (at least) one area where this development has not come as far as I think it should have: Unsecured content is still allowed in secure documents. Some clients do give the user a warning (MSIE for example, Opera and FF do not), but as far as I know none show a padlock for such pages.

Some may wonder why such mixing is a problem, so here is a short description of some issues:

  • Unsecured content can be read by anyone listening in the connection (Not very polite of them)
  • Even worse, somebody listening in is almost always in a position to change the content (But, that's ... criminal, right?)
  • Some content, like images, can tell the eavesdropper what you are doing.
  • Other content, like JavaScript and CSS, can directly change the content or appearance of the webpage. If modified by an attacker, the page could be made do things you neither you nor the designer would like, like send your password and credit card details off to the attacker.


Therefore, using unsecured content in an otherwise secure document can, in some cases, leak information, while in others it can compromise the security of the *entire* site.

A very "popular" excuse for using mixed security in a site is that it reduces load on the server, because of the cost of SSL connections. This excuse was debunked by Bob Lord (formerly of Mozilla) a few years ago. The most costly operation for a secure secure server is the private key decryption step of the initial SSL handshake, and on a well-configured server that is an operation done no more than once a day per client. The remainder of the extra work is the encryption, and modern encryption methods like AES have been designed to be as efficient as possible. Both of these operations can be performed using hardware accelerators, which should reduce your total hardware cost if you do have heavy traffic to your site.

By ensuring that your server is serving HTTP 1.1 content, that it properly support and specify cache directives, persistent connections and pipelining, that you can maximize your server's efficiency. If you still think your site is under too heavy a load, perhaps you can reduce the complexity of the documents, simplify images and scripting, and rely on CSS to place the content correctly?

In most cases where we see mixed-security content used these days, the reason is not cost reductions, it is due to copying code, for example an analytics JavaScript tag, without examination of the code, forgetting to change the URL to https, and during testing not noticing the security warnings some browsers do show (unless the testing is not done on a secure site, but an unsecure one, which might explain some of the curious aspects of this).

In my opinion such mixing is long past its use-by date. It may have been (or seemed) necessary in 1998, but not in 2010, especially with the much more hostile environment that now exists, such as phishing, malware, and the possibility of compromised WiFi networks.

During beta testing of IE7 Microsoft tried to block mixed-security content by default, but apparently found that "too many" sites broke, and backed down.

I wish Microsoft had not backed down. The "problem" with secure sites not working "properly" when blocking unsecured content will not disappear, and will have to be handled if/when blocking is reactivated. In my opinion, any such problems with broken sites will be short-lived, and the sites will probably be fixed within weeks of a policy shift.

One might hope that fewer sites are doing using mixed-security content when browsers start to block that type of problem, but I would not hold my breath.

However, I continue to see hotel booking sites, bank sites, and other important sites loading unsecured content as part of their secure pages, and even new sites do it.

In recent weeks I have noticed two high profile sites with such mixing of secure and unsecured content: Twitter and Citibank. In both cases the pages in question included external JavaScripts, which would allow an attacker to completely replace the content of the site, and, for example, to use it to steal passwords.

Twitter (contacted June 3rd) have recently been updating their site, including a secure site interface, but have not yet removed all the references to unsecured content on its main page. According to my information, they are working on it.

In <http://www.citi.com/> Citibank's case (contacted June 17th) there are two problems: Their main landing page is not encrypted, and the "Unknown page" (404) error page (e.g. the page for https://www.citibank.com/foo ) on one of their secure severs includes unsecured scripts, meaning that a MITM attacker can replace their main page (by editing the network traffic, for example in a WiFi-router) and is able to redirect the user to a server where he is (by replacing the unsecured script) in complete control over the content. The only way for a user to discover that he is under attack is to notice that there is no padlock and/or that the URL is not the one Citibank actually use for the service they are supposedly using, even though the URL belongs to Citibank.

I think that this is a big chink in the bank's security armor, but apparently Citibank does not think so, referring to the fact that the vulnerable server is not their main site any more (which does not matter as long as the server is online), that their services are secured with encryption (and that does not matter if the attacker can trick the user into believing they are on the right site) and that they have other security policies in place to help users in case their account is compromised. My impression is that they are not planning to fix the problem.

The problem with mixed-security sites will not disappear while they are allowed to work, so we do not gain anything by waiting, rather the opposite.

While I think blocking unsecured content in secure pages is the best solution, there are others who think that one should warn the user with a dialog, or display the page without any indication that it is partly loaded from a secure server, for example by removing the "https" in the URL displayed in the addressbar, or showing it as crossed out, because we "should allow sites to shoot themselves in the foot if they absolutely want to".

The problem with the warning dialog is that most users will usually just click through it to continue their business. Crossing out https is not much different from what clients do today, by not showing a padlock, and the problem with allowing sites to "shoot themselves in the foot" is that the foot actually belongs to the user.

In my opinion, blocking the mixing of secure and unsecured content is the best option, perhaps with an option for the user to manually trigger loading of the unsecured elements. It may break sites during a transition period, but such sites will be quickly fixed if the issue is critical for the site. The end result will be a more secure web.

In standards work, there is a suggestion being circulated which might help in this area: HTTP Strict Transport Security(a video of a presentation by the author is available here). HSTS is designed to let sites specify HTTPS as the only allowed connection method for their site or domain, amd it contains a suggestion to browser implementations that they block mixed-security content. However. it would only work for sites that ask for such a policy.

This is an area where few browser vendors, if any, can make a move on their own, but by joining forces we might be able to do something about the problem.

I therefore urge Adobe, Google, kHTML, Microsoft, Mozilla, Opera, and Safari, as well as other browser and framework vendors, to discuss a common solution to this issue, and agree on a time-line for when we will ship browsers and frameworks that implement that solution.



Previous articles on this topic:

Renego patched servers: A long-term interoperability time bomb brewing

, , , ...

As mentioned in a security group article a few weeks ago, over the past few months, we have been probing a large number of SSL/TLS servers to follow the adoption of servers that have fixed the TLS Renego vulnerability.

Over the past three months, about 12% of the tested servers have been updated to support the new TLS Extension that was developed to fix the issue. Extrapolating, and assuming the same growth rate, this means that it will take more than two years before "all" servers are patched, which in my opinion is much too long to leave a security vulnerability such as this unpatched.



At the same time, we have been observing a pattern that I think is of some long term concern: most of the servers that have been patched since early April are not fully TLS compliant. Specifically, these servers do not tolerate a client identifying its highest supported protocol version as 4.1 (a currently non-existent version; SSL 3 and TLS 1.x are using protocol version 3.x).



In the past few weeks, as many as 80-90% of the newly patched servers have refused to negotiate with our tester (the TLS Prober) when it claimed to support the hypothetical v4.1 TLS protocol version (or, as I call it, "TLS NG"). This is much higher than the 69% of all servers that generally exhibit the same problem.

The major TLS protocol version 4 is currently a hypothetical version of the protocol, and there are AFAIK no plans to write a specification that will use this major version number.

So, why worry?

I am concerned. If this version intolerance, which _is_ a violation of the TLS specification, is still in widespread use whenever protocol version 4 is defined, then we will, at best, have an interoperability problem; at worst we could have a serious security vulnerability.

Over the past 10+ years, TLS clients have had to cope with version-intolerant servers because older servers were not written to handle clients supporting newer versions than theirs of the protocol. It usually has been done by silently disabling the newer versions if the server does not tolerate them. This problem persisted well into TLS 1.0 deployment and was also extended to TLS Extension support, which also required clients to implement further fallbacks. This type of problem delayed Opera's activation of TLS 1.1 and TLS extensions for more than a year, after a scavenger hunt revealed the size of the problem, because we had to develop a way to handle the intolerant servers1,2.

These fallbacks are not just adding serious complications to our code (and every browser's code). They have the added potential to create security problems down the road, if (or more likely, when) a security problem develops in an older version of SSL or TLS that allows an attacker access to the protected data.

Therefore, it was very good that the TLS Renego RFC specifically reiterated the requirement that always existed about version and extension tolerance. Opera followed up on that by requiring Renego patched servers to tolerate a TLS 1.2 (version protocol 3.3) handshake, as mentioned in our article when we started testing.

So far, all servers in our list that have been updated with the Renego patch have implemented this properly with respect to SSL v3 and TLS 1.x tolerance. Very good!

However, it looks like some vendors unfortunately did not thoroughly think through what the version tolerance requirement in the TLS specifications really means. It does not mean, and has never meant, "We can refuse to negotiate with clients offering protocol version 4.0 or larger". It means "That client says it supports version 4.0 or higher, but we only support version 3.x, so we will only talk version 3.x with it".

In some cases, it seems that downstream vendors release updates that only include the Renego patch but did not pick the update that fixed the version intolerance problem. It might be that they did not think it was a security patch.

If these servers are still active when TLS NG (or whatever the next major version of SSL/TLS will be called) is defined and gets implemented, clients will either have to break these sites by refusing to connect to them, or we have to reintroduce the protocol fallback. As mentioned above, the fallback could create a security vulnerability.

Further, if a new major version of TLS is ever created, while there may be other reasons, it will most likely be for either one, or both, of the following two reasons:

  • There are new improved protocol techniques that are very incompatible with those currently used in TLS
  • The old system's cryptographic protection methods being discovered to have serious vulnerabilities, requiring a complete rewrite of the protocol


In the latter case, assuming the known problems are not so serious that support for older versions must be discontinued immediately, allowing an automatic fallback to be used would allow an attacker to trick a new client to talk to a new server using an old and _vulnerable_ protocol. Oops!

Therefore, since all servers are now being upgraded to protect against the Renego issue, we need to nip the v4-intolerance problem in the bud while it is still relatively small.

So far, we have been able to identify one vendor and have started to contact them about the issue. However, there is not yet a clear pattern to the server version information, which makes it very difficult to determine what other vendors are involved. It is also possible (probably very likely) that, based on the observed variation in server agent, the actual TLS servers in many cases are
SSL/TLS front-end accelerators or firewalls that do not directly inform the client about their involvement in the connection.

We will continue to attempt to identify vendors and contact them about this issue. We also have several other items being developed in relation to this, such as an online test utility.

Currently the *original* SSL/TLS implementations we know have been implemented with correct version tolerance are (some have not been release yet):

  • OpenSSL 0.9.8m (for cherry pickers, you can find the relevant patch here)
  • OpenSSL 1.0.0
  • NSS 3.12.6
  • Windows 7 (update not yet released, AFAIK, probably also applies to other windows versions)
  • RSA BSafe (version unknown, not sure if it has been released)


There may be other implementations which have not been included.

One thing we have discovered is that some customized variants of OpenSSL with the Renego patch do not include the above mentioned version tolerance patch. The maintainers of such derived distributions should include that patch in their codebase.

For other vendors who wonder whether they need to do anything to their Renego patched system, I may be able to help them if they contact me and provide a URL to a test server that I can test.

RFC for fixing the TLS Renego vulnerability published

, , , ...

As discussed in this Security Group article, a serious vulnerability in the SSL and TLS protocols renegotiation was discovered last year.

The update of the SSL and TLS protocol to fix the "Renego" vulnerability was published earlier today.

The RFC can be downloaded from http://www.rfc-editor.org/rfc/rfc5746.txt

As mentioned in the article, Opera 10.50 Beta 1 includes support for the updated protocol, although it is not fully activated yet due to usability reasons. In related news, Mozilla included support in their nightlies earlier this week.

American Express used revoked site certificate for weeks

, , , ...

In the past couple of weeks I have seen a few reports that the American Express site https://online.americanexpress.com/ (Testurl) was using a revoked certificate. This site redirects to various American Express services.

As of this morning (October 1st) the problem has been fixed.

As I have mentioned before, a revoked certificate is no longer valid, and therefore no browser should give access to the site. In the worst-case scenario, a certificate is revoked because the key has been compromised, and the site is being used by criminals for fraudulent purposes. Revocation of a certificate is analogous to blacklisting of a credit card because the card can be abused.

The available (human readable) information indicates that the current situation was not a worst-case scenario. What seems to have happened is that American Express requested the current certificate in May, then shortly thereafter requested an updated certificate, possibly to correct a minor mistake. The old certificate is revoked automatically after a reissue, and the Web site is expected to use the newer certificate, but it seems that American Express for some reason did not update their server. Since there is no way for a browser to distinguish between a benign mistake and a worst-case scenario; we have to assume the worst.

I am sure somebody is going to mention that the site worked without problems in many other browsers. Yes, it did, but that was because the certificate for this site is issued from a special Verisign certificate hierarchy (which is being retired) that does not specify the revocation list (CRL) download location, so browsers cannot go looking for the information, based on the certificate. The reason why the CRL is not specified may, as I understand it, have to do with the fact that the hierarchy was never intended to be used for Web site certificates, but for authentication of dedicated secure server-to-server connections. Opera may be one of just a few browsers that actually check the CRL for this hierarchy, but we can only do this because we have added the certificate to a list of overrides our repository that specifies where to fetch the CRL, otherwise we would not show a padlock on these sites (and as we discovered last year, Opera users really do check the padlock!).

A partially unanswered question is why reports appeared only so recently. Part of the reason is that a bug in our rootstore repository server caused the CRL override Opera uses to be dropped for 6-8 weeks (we fixed that two weeks ago), but that does not explain why there are no reports before the end of July. My guess, based on information from Netcraft, is that online.americanexpress.com is a new server that was brought online just a few months ago, which turns out to be well into the period when most of the CRL overrides were disabled by the repository bug, which means we did not check the CRL either for this site during that period o .

Once we learned of the problem, beginning of last week, we informed Verisign about it, and I assume they informed the American Express contact for this certificate about the problem.

Yet, more than a week later the revoked certificate was still being used by the Web site, until it finally got fixed last night. Frankly, I expected better from American Express.

Announcing the Security @ Opera home page

, ,

The Opera Security group just got its own home page.

"Security @ Opera" is where we will post information about security advisories, as well as occasional articles about security related topics.

The first article is about the recent developments concerning MD5 in SSL/TLS certificates.

W3C Web Security Context: User Interface Guidelines

, , , ...

The W3C's Web Security Context Working Group have just released the Last Call version of its "User Interface Guidelines" document, which is a set of recommendations for the security related UI in Web User Agents.

This specification deals with the trust decisions that users must make online, and with ways to support them in making safe and informed decisions where possible.



This document specifies user interactions with a goal toward making security usable, based on known best practice in this area. Subsequent testing of this specification will include conformance, interoperability, and usability testing.



If you want to comment on the document you are welcome to do so:

The W3C Membership and other interested parties are invited to review the document and send comments to public-usable-authentication@w3.org (with public archive) through 15 September 2008. We appreciate if comments follow these guidelines for writing good issues.




"Nobody checks the padlock" debunked by Opera users

, , , ...

There's been a number of criticisms directed at the padlock over the years, some of which may be correct, to some extent, at least.

One objection has been that users misunderstand the padlock's meaning, thinking it is an indicator of trustworthiness, rather than a protection rating for the connection.

Another objection (which flies in the face of the first), is that "nobody checks the padlock".

Well, I can't say much about the correctness of the first objection, but the past couple of weeks a growing number of Opera users have been working hard at debunking the second one.

In Opera 9.50 we added some new security features, and changed a some others. Two of these are the following:



Both of these turned out to encounter unexpected problems, not due to bugs in our implemenation, but at the CA side. In all cases the problem cause full validation of the certificates to fail, and as a result Opera reduces the security level of the connection so that a padlock will not be displayed. Quite a lot of people noticed.

For OCSP we encountered problems with at least three different brands of OCSP responders. These responders did not respond correctly when we sent a request for infromation about a certificate. It wasn't until a couple of weeks ago that the vendor for one of them discovered the reason for the problem, their expectation of input format was wrong. According to my information they have developed a patch, and I assume they are doing QA on it now, before sending it to the two (or more) CAs that are affected by this. The others have shown up recently and I do not have good information about the causes.

With CRLs we encountered two problems. Just before 9.50 was released we got the first reports about the first case, but it was not until a few days after the release that we got an overview of the situation.

It turns out that one specific CA created an intermediate certificate a few years ago for one of their certificate hierarchies. This certificate included an incorrect URL for where we should go to fetch the CRL, so when Opera fetches the CRL it gets a CRL created for a different hierarchy than the one being verified. As a result Opera's ceritificate validation code can't find the right CRL, and this step fails. This certificate became installed on thousands of servers before the mistake was discovered, and despite a campaign to replace the certificates, several hundred sites (many of them banks) still use the old certificate.

In another case, which came to light last week, another CA have issued an intermediate certificate directly below their root that does not contain a URL for a CRL, but the sub-ordinate certificate issued from that intermediate does specify a CRL. As Opera's certificate validation code expects that if one certificate have a CRL they all use CRLs, validation of the topmost intermediate certificate fails because it can't find the CRL for that certificate.

What can be done about this?

  • About the OCSP issue, we wait for the vendor patch, but in Opera 9.51 we have added an override that lets us use the certificate update system to specify which OCSP URLs that must use the old POST method. In addition one CA also deployed a workaround.
  • About the first CRL issue, the websites need to update their installed certificates, in the second case the CA must also issue a new updated certificate. While waiting for that, in 9.51 we are adding a similar override mechanism as for OCSP, by specifying extra CRL download locations for specific CAs. In the second case we might also be able to fix the issue if the CA has a root for that particular CA name by distributing it to all installations that access the affected sites (that is not a realistic option for the first case).


These changes will be active in the upcoming Opera 9.51 RC2 (Update: Now available), and is already active in the online certificate repository. You will get the updates the next time Opera checks for updates, or if you use Help->Check for updates.

Update July 9: There are some sites with certificates that does not provide any CRL or OCSP information in the site certificate, while the rest of the certificate in the chain have CRLs. Due to a minor bug, and restrictions in the override functionality, the override for that will not become active before 9.52 has been released. At present we are only aware of 10-20 sites affected by this.

More testing: Updated EV information and new Roots

, , , ...

A few hours ago the new online certificate repository that the most recent weeklies are using was updated with several new roots, and an additional CA, Comodo, was also provisionally EV-enabled

The new and the updated CAs are

  • America OnLine
  • Cisco
  • Comodo
  • Digicert


There is no need to download an updated Weekly (if you have one of the two recent ones). When you next restart one of the Opera 9.50 Weeklies with support for online certificate updates, it will immediately download the indexes, and download the new certificates when necessary. Please give it a minute to finish the update.

Here are a couple of testsites:


Known Issues: The complex certificate chain system used by Comodo encounters some, mostly hidden, problems with our OpenSSL certificate verification support, and that will cause some EV sites to not be recognized. We will try to fix it, but it may not be advisable to include a fix in 9.50.
[*] DigiCert:
[/LIST]

I do not currently have testcases for Cisco, as they have not yet started issuing certificates from the new root.

More about known issues:

  • We know there are some problems with OCSP and CRL responses (the two kinds of revocation information) from some Certificate Authorities. These problem may lead to the website getting a lower security level. We are looking into these problems together with the CAs. In last week's build some of the CRL problems will cause a "Fatal Error 50", in the most recent build that has been fixed. We may decide to work around some of these, but they should preferably be fixed by the CA.
  • At least one CA (who is not in our repository) is using CRLs with a critical extension, which will cause the secure connection to fail with error code 554. In this case we are following the standard, although one might wonder why the specification says "Although the extension is critical, conforming implementations are not required to support this extension". The problem have been "fixed" internally by recognzing the extension, then ignore it, as we do not need it.
.
February 2012
S M T W T F S
January 2012March 2012
1 2 3 4
5 6 7 8 9 10 11
12 13 14 15 16 17 18
19 20 21 22 23 24 25
26 27 28 29