We have found a privacy/security bug in the Java implementation of the Netscape Navigator. It is very easily possible for an applet to find out the pathname of the directory in which the Netscape Navigator was started. This information could then be sent back to a CGI program for logging. Clearly this information should not be available to an applet, as is indicated by the fact that applets are prevented from reading the "user.home" and "user.dir" system properties. When the Netscape Navigator is run under the Windows 95 OS, the pathname usually does not contain any critical information. However, when the Navigator is run under a multi-user network OS, such as UNIX, the pathname often contains the e-mail and/or login name of the user. In addition, the pathname might reveal details about the topology of the user's network, which an experienced hacker might be able to exploit. There are two ways to protect yourself from this problem: Either start up the Netscape Navigator in a directory whose pathname does not reveal any critical information, or disable Java altogether (Options | Security Preferences | General). A system administrator can protect his network by configuring the HTTP proxy server not to retrieve Java ".class" files. This bug is present in at least the following versions of the Navigator: 2.0 2.01 3.0b2 2.0GoldB1 2.01Gold and in the implementations for at least the following platforms: SunOS 4.1.2, 4.1.3, 4.1.4 SunOS 5.3, 5.4, 5.5 Windows 95, Windows NT IRIX 5.2, 5.3 HP-UX A.0903, A.0905 Linux 1.2.10, 1.2.13 FreeBSD 2.1.0-RELEASE OSF1 V3.2 We have not tested whether this bug also exists in Sun's HotJava browser. We will release full details of the bug as soon as Sun and Netscape have issued patches which fix the problem. Full details have been sent to Sun and Netscape. This announcements has also been posted to the "comp.lang.java" newsgroup and has been sent to CERT. Daniel Abplanalp and Stephan Goldstein (firstname.lastname@example.org) Berne, Switzerland
It is well known that numerous computer programs are so poorly designed and implemented that they cannot handle special cases of people's names (e.g., the case of the person simply named "Smith", as in RISKS-18.05). In some other cases, there are real physical (and other) constraints that are hard to code around, such as in the case of some colleagues whose long names risk falling off the edge of their company badges. Sometimes people just push things to far to the extreme. The Guardian Weekly reported (in the issue ending 21 Apr 1996, p.4) that "a Swedish court has fined a couple $660 for breaking the law by naming their son Brfxxxcccxxmnnpcccclllmmnprxxvvclmnckssqlbb11116--or Albin for short." Does anyone know what law was broken, and can anyone decode the meaning or origin, if any, of this choice of name? Li Gong, SRI Computer Science Laboratory, http://www.csl.sri.com/~gong [Their son will be lucky if he does not get called Albin0, even the full given name might appear to be white noise. This is clearly job for a new Sesame Street song to help us remember the given name. PGN]
Baltimore just finished creating a brand new technologically advanced "central booking" building where police take people after they've been arrested. Unfortunately, this heavily computerized system has become media fodder lately as people get lost in the building and are not released for days. I've seen one television news report that interviewed a woman who said she spent five days in the building because of a minor warrant for an unpaid fine. Bail was posted several times and lost. The Thursday 18 Apr 1996 edition of the *Baltimore Sun* reports that the system is now working faster after the police started overriding the computers. For instance, there was an automated system by which prisoners could be called up for their hearings. The handheld computers that carried these messages, however, didn't work and now the City has detailed five extra officers to escort the prisoners instead. Another woman complained that her 18-year-old retarded child was held a full day after bail was set. She says that she was speaking with booking center over the telephone when the child knocked at the door. The booking center was telling her that her son could *not* be released until all of the computer records were complete. I can only hope that the same glitch won't let out a dangerous criminal. The article ends by noting that the District Court Commissioner says that "Anybody [who's] been in this building would be a damn fool to [go] back to it."
I would like to bring your attention to some recent joint work by Advanced Micro Devices, Inc., and Computational Logic, Inc., in which the ACL2 theorem prover was used to prove the correctness of an algorithm of commercial interest. In particular we proved the correctness of the kernel of the floating-point division algorithm on the AMD5K86, the first Pentium-class processor produced by AMD. Roughly speaking, we proved that the algorithm implements division on the double extended precision normal and denormal numbers of the IEEE standard, in the sense that (under appropriate hypotheses) it returns the floating-point number obtained by rounding the ``infinitely precise'' quotient by the method and to the precision specified by a given rounding mode. The permitted rounding modes include round to 0, round away from 0, round to nearest, round to positive (or negative) infinity, and ``sticky'' rounding. The proof is quite interesting, involving as it does the formalization of a lot of floating-point ``folklore'' and classical numerical analysis. J Strother Moore The paper may be obtained via the URL: http://devil.ece.utexas.edu:80/~lynch/divide/divide.html The title and abstract are shown below. A Mechanically Checked Proof of the Correctness of the Kernel of the AMD5K86 (tm) Floating-point Division Algorithm J Strother Moore (Moore@cli.com) Tom Lynch (Tom.Lynch@amd.com) Matt Kaufmann (Matt_Kaufmann@email.mot.com) ABSTRACT: We describe a mechanically checked proof of the correctness of the kernel of the floating-point division algorithm used on the AMD5K86 microprocessor. The kernel is a non-restoring division algorithm that computes the floating-point quotient of two double extended precision floating-point numbers, p and d (d \= 0), with respect to a rounding mode, mode. The algorithm is defined in terms of floating-point addition and multiplication. First, two Newton-Raphson iterations are used to compute a floating-point approximation of the reciprocal of d. The result is used to compute four floating-point quotient digits in the 24,,17 format (24 bits of precision and 17 bit exponents) which are then summed using appropriate rounding modes. We prove that if p and d are 64,,15 (possibly denormal) floating-point numbers, d \= 0 and mode specifies one of six rounding procedures and a desired precision 0 < n <= 64, then the output of the algorithm is p/d rounded according to mode. We prove that every intermediate result is a floating-point number in the format required by the resources allocated to it. Our claims have been mechanically checked using the ACL2 theorem prover.
Taking advantage of yet another incentive offer, I recently switched my long distance carrier to MCI. They sent me the standard yet-another-piece-of-plastic-to-stick-in-my-wallet calling cards. The way these cards work is that you call an 1-800 number and type in your code consisting of your phone number followed by your PIN (Personal Identification Number) which happens to be printed on the card. Enclosed with the cards was a piece of paper in which MCI wisely suggests that you change your PIN to something other than what they assigned to you and printed on the card: Customizing your PIN Choosing your own four-digit number is the best way to assure you'll never forget your PIN. Make it the month and year of a loved one's birthday or use the same password you have for your voice mail or computer. We'll quickly replace the PIN we assigned you with any four digits you choose - just call 1-800-476-7306 For some strange reason MCI is recommending you to do exactly the opposite of what good security practices would proscribe! Not only do they suggest that you use an easily-breakable password such as an important date, but they recommend a practice that would weaken the security of potentially more sensitive information in a voice-mail or computer system. Of course, what probably prompted note from MCI was a desire to prevent MCI's customer service department from being inundated with calls from people who forgot their PINs. This alludes to the associated risk of requiring people to remember Yet Another Password (YAP). -chad
I find it necessary every few weeks to set the clock on my Unix box at home because PCs are not generally known for incorporating accurate timekeeping hardware. Shortly before doing this recently, I ran across this notice on the web page of the US Naval Observatory Directorate of Time, operates a number of stratum-1 NTP servers, and is otherwise the official source of time in the United States. http://tycho.usno.navy.mil/phonecheck.html On the morning of April 16, the Directorate of Time moved its Master Clock #1 Hydrogen Maser to a new environmental chamber in Washington, D.C. Timing operations were switched to Master Clock #2 Hydrogen Maser. All except our network time servers tick and tock, which failed to switch over, and jumped back 744.126 seconds between 12:32:58 UT and 14:23:46 UT, when the problem was corrected. This jump was 10 million times the normal precision of these systems. Presumably, other stratum 1 NTP servers operating from local clocks, WWV or GPS broadcasts successfully made both changeovers and never noticed. Many protocols, kerberos for instance, would not take kindly to a ten-minute drift. While the close coupling of hosts "tick" and "tock" to the actual US time standard make them appealing servers as far as nerdly bragging rights are concerned, in this instance it also made them vulnerable to the _process_ by which US time is determined. Dave Hsu <email@example.com> Systems Programmer Software Development Group UUNET Technologies http://www.va.pubnix.com
I have found a serious bug (feature?) in Windows 95. It does not properly treat files that contain certain characters in their names. These characters include those with ASCII values of 255, 254, 249, 244, 23* and some others, all above 127. I have not found a common rule (so I probably failed the Microsoft IQ test :). This problem was noted by Olcay Cirit in RISKS-17.64 in regards to ASCII 229. He wrote that if this character is the first in a filename, then the file cannot be deleted, copied, renamed or executed. Actually, *any* of these characters at *any* place in the filename will spoil the file. Such files are visible in Explorer, with bad characters shown as underscores. You can create shortcuts to them, view their properties or even try to rename them. But any serious operation has to be performed from the command line. A pecularity of my DOS file managing program is that it uses direct disk access to delete files. I could not do that under Windows 95 until recently, so to deal with the "bad characters" I had to reboot into DOS prompt. Then I discovered the the "LOCK" command will allow DOS utilities to access disk directly. However, this is probably an undocumented command, it's absent in the DOS help file, and it is not an executable file. A note for those unfamiliar with the "funny" characters. You can enter any character at the DOS prompt by holding the Alt key and pressing the keys with digits for the character at the numeric keypad. For example, Alt+097 is 'a', and Alt+255 is one of the "bad characters". So, to see the described behaviour for yourself, create at the DOS prompt a new file that has a "bad character" in its name, then try to do something with it in Windows 95. The RISKS? Aside from user confusion and possible pranks (which cannot do much harm, because you can always go to the command line and fix the things), there is another issue. Usually filenames with "bad characters" are not used. However, here in Russia one way of russifying a PC is to replace all those Greek, German and Swedish symbols that reside in the upper part of the ASCII table with Russian letters. So, if a Russian user had many files with Russian names, and then switched to Windows 95... Surprise, surprise! You can't manage your old files there. I have to confess that I do not know how Russian edition of Windows 95 works. I am using the US edition. Simon (Vsevolod Ilyushchenko) firstname.lastname@example.org email@example.com
Bell Atlantic NYNEX Mobile has a web page at: http://www.banm.com >From that web page I got an e-mail address (firstname.lastname@example.org) and sent a report of a service difficulty. I also identified my account and the dealer from which I signed up for service. Fortunately I did NOT post my PIN number in the message. The dealer then called me to say that my message had gone "all over the internet"! I then e-mailed BANM again asking: "Do private e-mail messages such as this wind up being spread all over the internet by your site? I was not aware of this. Why don't you warn people sending you e-mail messages that this happens. I would assume that some of your customers would feel that this is an invasion of privacy and possibly litigatable. BELL ATLANTIC NYNEX MOBILE has not replied yet. I think that this should be a warning to people sending BANM what they THINK to be private e-mail messages to be advised that they may wind being published on their web page. So be sure not to put confidential information in messages to BANM. Ray Normandeau, email@example.com http://www.buzznyc.com/actors/res.normandeau.raymond.html
>Anatoly Voronov, the director of Glasnet, an Internet service provider in >Russia, says: "It is just incredible when I hear people talking about how >open the Web is. It is the ultimate act of intellectual colonialism. The >product comes from America so we either must adapt to English or stop using >it. ... Sigh. As always clueless get most publicity. Apparently Mr. Director of small but noisy local service provider is not aware that Russia-language newsgroup traffic is second to English-language. Apparently he's not aware that domestic ISPs in Russia offer quite a lot of Russian-language material. Apparently he does not even realize that understanding of Internet technology in Russia is good enough that there happened to be quite a few recent Russian immigrants taking senior positions in U.S. telecom industry. The prevalence of English is merely due to the need for some common language. A modern man cannot be considered educated if he cannot read it. It's like Latin centuries ago. How can anybody's head to be so screwed as to consider the only real advance in destroying the inter-cultural communication barriers to be "the ultimate act of intellectual colonialism" (ain't that an oxymoron?) is beyond my comprehension. Maybe French government should send him a job offer. --vadim
1) Nicholas (sp?) Negroponte of MIT's Media Lab begins his book "Being Digital" by emphasizing the difference between transporting bits and transporting atoms. The Internet moves bits. The Soviet Union, in its heyday of real rather than intellectual colonialism, moved atoms, i.e., people, in very large numbers, specifically natives out of subordinated lands, Russians into them. 2) The "product", I thought, came largely from *CERN*: an *international* organization, located as it happens in *France*. [...]
I think one of the most frightening aspects of this was the penultimate screen image. There is an option "YES" (i.e., proceed) but none for "NO" (i.e., I've changed my mind). OK, the very last screen _does_ give you this choice, but ... how is the intended patient to know that there's another screen coming up (unless they go through a dress rehearsal)??? That screen yelled out at me from the (I think it was last Tuesday's Australian newspaper but could have been the Sun's) Computer Section page. There was no obvious way to change one's mind. This is the ultimate bad GUI design - no apparent option to cancel or go for a "none of the above" decision. Stupidity? Criminal? I don't know, but it was downright frightening. It reminds me of (and I don't mean to belittle the thread) dialog boxes (or alerts) which used to notify you of a (serious) System Error on the Mac. It gave you the "choice" of doing a restart (reboot) or a quit. The trouble was that the quit almost always never worked and the thing froze anyway. Current Mac SW gives you the "choice" of restarting only (apart from those that automatically quit to the Finder). Perhaps the subtle differences between an Alert Where you usually don't have a choice - e.g., "You are about to die.. Press OK to continue" and all other forms of dialog boxes Where you _do_ have a choice (and even an escape hatch not covered by those being solicited).. need to be made more obvious. The ornamentation (at least on the Mac) for both types of displays (windows) are very similar. The windows in question (from memory) actually seemed more like web browser pages with the YES/NO options as links. As for an escape hatch, there should always be one, regardless. It should never be out of view (ie, it should *never* be scrolled off the top). Say, for example, a patient had a change of heart with that penultimate window displayed. OK, there's always the "quit" menu option, hitting the power switch, etc, but how comfortable and certain is he/she that this hasn't left the drip (or whatever) still in a primed state? There is no confirmation that equipment has been de-activated, especially if he/she hits the power/restart switch. Critical systems indeed. Paul Menon, Dept of Computer Science, Royal Melbourne Institute of Technology, 124 Latrobe Street, Melbourne 3001, Victoria, Australia +61 3 9660 3209/2348 [Various other folks commented on other aspects of this situation, testability using placebos, social implications, the long-standing RISKS issue of trying to solve social problems with technological solutions, etc. All very interesting, but drifting a little too much. TNX. PGN]
In: Risks-Forum Digest Friday 19 April 1996 Volume 18 : Issue 05 >From: Edupage Editors <firstname.lastname@example.org> >Subject: More Microsoft Viruses (Edupage, 16 Apr 1996) >First there was the Word virus — now there's a Word Prank Macro Virus, Before Microsoft admitted it *was* a virus, their term for WordMacro.Concept (the "original"), was the Word Prank Macro. This isn't new at all. >located in a document on ActiveVRML, Microsoft's software tool for >developing 3-D Web sites. But what's worse, is that Microsoft had to inform >the programmers who attended its Professional Developers Conference last >month that one of the CD-ROMs it distributed was infected. A cure is posted Concept is everywhere. Of those who work in large corporations, the only ones who *don't* regale me with stories of massive infestations are those who do not know how to check for it. Microsoft has been a repeat offender: we are constantly hearing of new disks and CD-ROMs from MS that contain infected documents. >on Microsoft's Web site < http://www.microsoft.com/ > (*Investor's Business >Daily*, 15 Apr 1996, A8) Microsoft's anti-WordMacro/Prank/Concept package is a Word macro itself. It takes a piecemeal approach, and has been updated several times as new macro viruses have been discovered. The protection provided has holes: it is best not to rely on it. Edupage has had a large number of erroneous virus reports over the past year. No worse than the general press, of course (where their material is obtained), but somewhat disturbing in the technical community they distribute to. email@example.com firstname.lastname@example.org email@example.com
I have spoken to several people who have signed up for the new toll-free US 888 numbers (addition to the older 800 numbers). They are seeing a precursor to the year 2000 problem - many systems don't recognize 888 numbers. One story was told that a sales person called his 888 numbered voice mail system from his hotel room. When he went to check out, he had a bill for over $60 in phone charges. It seems that the local hotel did not recognize the prefix 888, and so assumed it was a long distance call and charged the highest toll rates. Other stories tell of pay phones that will not accept calls using 888, car phones that reject calls dialed with 888, as well as a myriad of other system related glitches. The change seems simple - just tell your system that 888 is a valid toll free prefix. However, as with the Y2K problem, it seems that plenty of systems either were not ready for the change, or are incapable of accepting it. If this is any indication of the Y2K RISK, I'm moving to a remote mountain top in 1999.
Does someone have ideas and suggestions for alternatives to the Social Security number for large organizations with large databases - methodologies like Alpha Search or Soundex? Are there other ways to manage huge databases without any use of SSNs or other numerical identifiers? Robert Ellis Smith, Privacy Journal, Providence RI 401/274-7861 [Respond to Bob, please, and let's see if he can summarize for RISKS. PGN]
Please report problems with the web pages to the maintainer