SHAKE XOFs
Hello, I noticed that the OASIS draft for extending PKCS#11 with SHA-3 also specifies new Mechanisms for SHAKE128/256. They introduce them as Key Derivation functions. I wonder if this would also be the way to introduce this into JCA, at the moment XOFs have been a non-goal of JEP287, but there is some use for them In modern protocols I would guess. (This request was inspired by a discussion on the bouncycastle crypto-dev mailing list about missing algorithms for it). Gruss Bernd -- http://bernd.eckenfels.net
Re: SHAKE XOFs
On 4/11/2018 5:37 AM, Bernd Eckenfels wrote: Hello, I noticed that the OASIS draft for extending PKCS#11 with SHA-3 also specifies new Mechanisms for SHAKE128/256. They introduce them as Key Derivation functions. Interesting. Though to be pedantic, it looks like they introduce key derivation mechanisms that are based on SHAKE128/256. I wonder if this would also be the way to introduce this into JCA, at the moment XOFs have been a non-goal of JEP287, but there is some use for them In modern protocols I would guess. (This request was inspired by a discussion on the bouncycastle crypto-dev mailing list about missing algorithms for it). Continuing the pedantry, it would be reasonable to put these SHAKE128/256-based-KDFs under the KDF API (once that API exists). But the underlying SHAKE XOFs probably belong in a different API like MessageDigest or a new API that is more appropriate for XOFs. I expect that adding the XOFs to the API will be non-trivial because we don't have an obviously good place to put them. I think it would be fine to put them in MessageDigest, but we would need a way to specify the output length. We will need SHAKE256 for Ed448[1], but my initial thought was to do a private implementation, because I don't know if these functions are useful enough to justify the effort of the API design. Maybe we can make an API for them as a separate effort. It's also worth noting that the (bare) XOFs are not very good KDFs because they allow key extraction through related output attacks. [1] https://bugs.openjdk.java.net/browse/JDK-8187789
Re: RFR: ChaCha20 and ChaCha20/Poly1305 Cipher implementations
Yes, that does appear to be the case, good catch! I'll make that change. --Jamil On 4/11/2018 7:18 AM, Thomas Lußnig wrote: Hi, i found another point. The "key" field can be removed from ChaCha20Cipher. 1) This field is only set once and later checked if it was initialized. But we do not want to knew is the key exists but if key bytes exists. 2) So if two lines are changed from key to keyBytes we can remove this unused field. Gruß Thomas http://cr.openjdk.java.net/~jnimeh/reviews/8153028/webrev.02/src/java.base/share/classes/com/sun/crypto/provider/ChaCha20Cipher.java.html Lines 426 , 461 (change to keyBytes) if (key == null Line 75+507 (remove) private Key key; this.key = key; On 4/11/2018 12:34 AM, Jamil Nimeh wrote: Hello everyone, This is a quick update to the previous webrev: * When using the form of engineInit that does only takes op, key and random, the nonce will always be random even if the random parameter is null. A default instance of SecureRandom will be used to create the nonce in this case, instead of all zeroes. * Unused debug code was removed from the ChaCha20Cipher.java file * ChaCha20Parameters.engineToString no longer obtains the line separator from a System property directly. It calls System.lineSeparator() similar to how other AlgorithmParameter classes in com.sun.crypto.provider do it. http://cr.openjdk.java.net/~jnimeh/reviews/8153028/webrev.02/ Thanks, --Jamil On 03/26/2018 12:08 PM, Jamil Nimeh wrote: Hello all, This is a request for review for the ChaCha20 and ChaCha20-Poly1305 cipher implementations. Links to the webrev and the JEP which outlines the characteristics and behavior of the ciphers are listed below. http://cr.openjdk.java.net/~jnimeh/reviews/8153028/webrev.01/ http://openjdk.java.net/jeps/329 Thanks, --Jamil
Re: RFR: ChaCha20 and ChaCha20/Poly1305 Cipher implementations
Okay, I will check that out and fix as appropriate. Thanks for the comments! --Jamil On 4/11/2018 10:56 AM, Thomas Lußnig wrote: Hi, same with key/keyBytes is with the class Poly1305 also here the key is not used. Gruß Thomas On 4/11/2018 5:54 PM, Jamil Nimeh wrote: Yes, that does appear to be the case, good catch! I'll make that change. --Jamil On 4/11/2018 7:18 AM, Thomas Lußnig wrote: Hi, i found another point. The "key" field can be removed from ChaCha20Cipher. 1) This field is only set once and later checked if it was initialized. But we do not want to knew is the key exists but if key bytes exists. 2) So if two lines are changed from key to keyBytes we can remove this unused field. Gruß Thomas http://cr.openjdk.java.net/~jnimeh/reviews/8153028/webrev.02/src/java.base/share/classes/com/sun/crypto/provider/ChaCha20Cipher.java.html Lines 426 , 461 (change to keyBytes) if (key == null Line 75+507 (remove) private Key key; this.key = key; On 4/11/2018 12:34 AM, Jamil Nimeh wrote: Hello everyone, This is a quick update to the previous webrev: * When using the form of engineInit that does only takes op, key and random, the nonce will always be random even if the random parameter is null. A default instance of SecureRandom will be used to create the nonce in this case, instead of all zeroes. * Unused debug code was removed from the ChaCha20Cipher.java file * ChaCha20Parameters.engineToString no longer obtains the line separator from a System property directly. It calls System.lineSeparator() similar to how other AlgorithmParameter classes in com.sun.crypto.provider do it. http://cr.openjdk.java.net/~jnimeh/reviews/8153028/webrev.02/ Thanks, --Jamil On 03/26/2018 12:08 PM, Jamil Nimeh wrote: Hello all, This is a request for review for the ChaCha20 and ChaCha20-Poly1305 cipher implementations. Links to the webrev and the JEP which outlines the characteristics and behavior of the ciphers are listed below. http://cr.openjdk.java.net/~jnimeh/reviews/8153028/webrev.01/ http://openjdk.java.net/jeps/329 Thanks, --Jamil
RFR JDK-8029661: JDK-Support TLS v1.2 algorithm in SunPKCS11 provider
Hi Valerie, I have an important customer awaiting this change. Can you please provide a status update as to the progress of this effort and any information available as to when it may be completed. Thanks, Jack
Re: RFR: ChaCha20 and ChaCha20/Poly1305 Cipher implementations
Hi, i found another point. The "key" field can be removed from ChaCha20Cipher. 1) This field is only set once and later checked if it was initialized. But we do not want to knew is the key exists but if key bytes exists. 2) So if two lines are changed from key to keyBytes we can remove this unused field. Gruß Thomas http://cr.openjdk.java.net/~jnimeh/reviews/8153028/webrev.02/src/java.base/share/classes/com/sun/crypto/provider/ChaCha20Cipher.java.html Lines 426 , 461 (change to keyBytes) if (key == null Line 75+507 (remove) private Key key; this.key = key; On 4/11/2018 12:34 AM, Jamil Nimeh wrote: Hello everyone, This is a quick update to the previous webrev: * When using the form of engineInit that does only takes op, key and random, the nonce will always be random even if the random parameter is null. A default instance of SecureRandom will be used to create the nonce in this case, instead of all zeroes. * Unused debug code was removed from the ChaCha20Cipher.java file * ChaCha20Parameters.engineToString no longer obtains the line separator from a System property directly. It calls System.lineSeparator() similar to how other AlgorithmParameter classes in com.sun.crypto.provider do it. http://cr.openjdk.java.net/~jnimeh/reviews/8153028/webrev.02/ Thanks, --Jamil On 03/26/2018 12:08 PM, Jamil Nimeh wrote: Hello all, This is a request for review for the ChaCha20 and ChaCha20-Poly1305 cipher implementations. Links to the webrev and the JEP which outlines the characteristics and behavior of the ciphers are listed below. http://cr.openjdk.java.net/~jnimeh/reviews/8153028/webrev.01/ http://openjdk.java.net/jeps/329 Thanks, --Jamil
Re: RFR: ChaCha20 and ChaCha20/Poly1305 Cipher implementations
Hi, ok that is an good point that if we have more values in the structure than we use this can make some confusion. I was only looking from the coding point of view and saw that if i can use the same structure for booth cases this can reduce the coding overhead. But i can fully understand your point. Gruß Thomas On 4/10/2018 11:37 PM, Jamil Nimeh wrote: Hello Thomas, et al., On 3/26/2018 1:49 PM, Jamil Nimeh wrote: Hi Thomas, thanks for the feedback 1. Were there guidelines? Not really, though I looked at other parameter definitions in com.sun.crypto.provider and tried to follow along the same lines that they do. One thing that should be changed is the LINE_SEP assignment shouldn't be an explicit getProperty call. I noticed most are doing System.lineSeparator() so I'll change my implementation to match that. None of these params appear to stringify as json, so I'll probably keep things consistent with the other parameter output. 2. You make a fair point with respect to a null SecureRandom. I can make that spec change. 3. Let me think on this one - I shied away from ChaCha20ParameterSpec for AEAD mode only because you have this nonce field that is set but gets ignored. But making ChaCha20ParameterSpec an IvParameterSpec potentially runs into the same issue were it used for a ChaCha20-Poly1305 cipher. If I had to choose between the two I think I'd go with allowing ChaCha20ParameterSpec to be used with CC20-P1305 rather than making it a subclass of IvParameterSpec. Doing the former helps from a type safety perspective that you couldn't use a ChaCha20ParameterSpec with other Ciphers that require an IvParameterSpec. I know I had some discussions early on in the design where we talked about this, I need to refresh my memory as to why we didn't allow it. Finally getting back to #3. Took me a while to find early discussions on this. The primary objection to ChaCha20ParameterSpec being used with ChaCha20-Poly1305 (as opposed to plain old ChaCha20) has to do with the configurable block counter. You have this parameter that is not used, and consumption of this type of AlgorithmParameterSpec then leaves it to documentation to define what happens (is it ignored? Used despite what the spec says? Set to some default value regardless of what the caller sets there?). Using IvParameterSpec with ChaCha20-Poly1305 is more clear because it only allows the caller what they need to get CC20/P1305 going, the nonce. Respectfully, I would like to keep this as-is. --Jamil On 3/26/2018 12:45 PM, Thomas Lußnig wrote: Hi Jamil, 1) where there any guidelines about how the engineToString should be formatted ? I ask because i wondering why we need two new lines with access to the System property. If it is represented as single line json no need to line break would be needed. Gruß Thomas /** * Creates a formatted string describing the parameters. * * @return a string representation of the ChaCha20 parameters. */ @Override protected String engineToString() { String LINE_SEP = System.getProperty("line.separator"); HexDumpEncoder encoder = new HexDumpEncoder(); StringBuilder sb = new StringBuilder(LINE_SEP + "nonce:" + LINE_SEP + "[" + encoder.encodeBuffer(nonce) + "]"); return sb.toString(); } 2) I do not think it is an good idea to say no secureRandom=null will cause IV to be null. I see here the risk of weak implementations. I would suggest to throw an Exception to enforce secure usages. If someone really want an insecure IV he can provide am SecureRandom implementation retuning 0 only or an matching IV. * @param random a {@code SecureRandom} implementation. If {@code null} * is used for the random object, then a nonce consisting of all * zero bytes will be used. Otherwise a random nonce will be * used. 3) If ChaCha20ParameterSpec would extends IvParameterSpec if would be valid for booth modes in engineInit. Even if the counter is not needed. As an alternative i would allow ChaCha20ParameterSpec also for AEAD mode. Grup Thomas On 3/26/2018 9:08 PM, Jamil Nimeh wrote: Hello all, This is a request for review for the ChaCha20 and ChaCha20-Poly1305 cipher implementations. Links to the webrev and the JEP which outlines the characteristics and behavior of the ciphers are listed below. http://cr.openjdk.java.net/~jnimeh/reviews/8153028/webrev.01/ http://openjdk.java.net/jeps/329 Thanks, --Jamil
Re: RFR: ChaCha20 and ChaCha20/Poly1305 Cipher implementations
Hi, same with key/keyBytes is with the class Poly1305 also here the key is not used. Gruß Thomas On 4/11/2018 5:54 PM, Jamil Nimeh wrote: Yes, that does appear to be the case, good catch! I'll make that change. --Jamil On 4/11/2018 7:18 AM, Thomas Lußnig wrote: Hi, i found another point. The "key" field can be removed from ChaCha20Cipher. 1) This field is only set once and later checked if it was initialized. But we do not want to knew is the key exists but if key bytes exists. 2) So if two lines are changed from key to keyBytes we can remove this unused field. Gruß Thomas http://cr.openjdk.java.net/~jnimeh/reviews/8153028/webrev.02/src/java.base/share/classes/com/sun/crypto/provider/ChaCha20Cipher.java.html Lines 426 , 461 (change to keyBytes) if (key == null Line 75+507 (remove) private Key key; this.key = key; On 4/11/2018 12:34 AM, Jamil Nimeh wrote: Hello everyone, This is a quick update to the previous webrev: * When using the form of engineInit that does only takes op, key and random, the nonce will always be random even if the random parameter is null. A default instance of SecureRandom will be used to create the nonce in this case, instead of all zeroes. * Unused debug code was removed from the ChaCha20Cipher.java file * ChaCha20Parameters.engineToString no longer obtains the line separator from a System property directly. It calls System.lineSeparator() similar to how other AlgorithmParameter classes in com.sun.crypto.provider do it. http://cr.openjdk.java.net/~jnimeh/reviews/8153028/webrev.02/ Thanks, --Jamil On 03/26/2018 12:08 PM, Jamil Nimeh wrote: Hello all, This is a request for review for the ChaCha20 and ChaCha20-Poly1305 cipher implementations. Links to the webrev and the JEP which outlines the characteristics and behavior of the ciphers are listed below. http://cr.openjdk.java.net/~jnimeh/reviews/8153028/webrev.01/ http://openjdk.java.net/jeps/329 Thanks, --Jamil
Re: RFR 8200468: Port the native GSS-API bridge to Windows
Hi Valerie I updated the webrev at http://cr.openjdk.java.net/~weijun/8200468/webrev.01/ The only change is that I prepend "GSS_DLLIMP" to all gss_* functions in gssapi.h. The file has the following lines 283 #if defined (_WIN32) && defined (_MSC_VER) 284 # ifdef GSS_DLL_FILE 285 # define GSS_DLLIMP __declspec(dllexport) 286 # else 287 # define GSS_DLLIMP __declspec(dllimport) 288 # endif 289 #else 290 # define GSS_DLLIMP 291 #endif I added it so the exact same header file can be used to write a native GSS-API library which would export these functions. Is this OK? Tests run fine with both MIT krb5 and Heimdal libraries. Thanks Max > On Apr 4, 2018, at 10:19 AM, Weijun Wangwrote: > > Hi All > > Please take a review at > > http://cr.openjdk.java.net/~weijun/8200468/webrev.00/ > > Like in *nix, native GSS-API bridge is turned on by setting > -Dsun.security.jgss.native=true. Please note there is no default native > GSS-API library on Windows and you need to supply your own, like this: > > java -Dsun.security.jgss.native=true > -Dsun.security.jgss.lib=/path/to/gssapi64.dll App ... > > You can manually test the change with > > jtreg -Dnative.krb5.libs=j=,n=/path/to/gssapi64.dll > test/jdk/sun/security/krb5/auto/BasicProc.java > > Thanks > Max > > p.s. You can get a gssapi64.dll from > https://web.mit.edu/KERBEROS/kfw-4.1/kfw-4.1.html.