According to reports I heard on the radio this morning, a US Airways Boeing 737 came less than 100 feet (some reports say withing 50 feet) landing on smack on top of a small corporate plane at LaGuardia Airport in New York City on Wednesday, 2 Dec 1998. The passenger flight was given permission to land, despite the fact that the smaller plane was waiting to take off on the same runway. The near-disaster has been blamed on too few controllers being too distracted. Dave Weingart, Strategix Solutions firstname.lastname@example.org 1-516-682-1470
I found the following item in the *Los Angeles Times* 16 Dec 1998. Especially intriguing are the spokesman's comments. There is also the nagging question of why there is an operator on a fully automated system in the first place. ...there IS a runaway train on Capitol Hill. The automatic brakes on the Senate subway between the Russell Office Building and the Capitol failed last week, sending the train crashing into a wall and slightly injuring the operator and the two other people on board. In the best congressional spirit, a spokesman for the architect of the Capitol stressed that "there was no operator fault involved. It's all automatic," said Herb Franklin, "and it's supposed to stop by itself." Thomas A. Russ, USC/Information Sciences Institute email@example.com
My office's frame relay link to our corporate WAN was out for a day on 26 Nov 1998. As it turns out, we were not alone. A railway backhoe operator accidentally cut an AT&T Canada fibre cable along the rail line between Toronto and Windsor, crashing computers, knocking out phone lines, and generally disrupting communications in southern Ontario. The main branch of the Bank of Nova Scotia was computerless. Rerouting of Internet lines via the U.S. was slow because of the Thanksgiving holiday. [Source: Veronique Mandal, ``Cut Cable Paralyses Network - Severed line brings hi-tech to its knees'', *Windsor Star, London Free Press, Chatham Daily News*, 27 Nov 1998; PGN Abstracting]
From The Times, 9 Dec 1998, p2: Bug threat to missile The Ministry of Defense admitted for the first time that the millennium bug could have left Britain vulnerable to air attack. It discovered that the Rapier anti-aircraft missile would have failed to retaliate. The problem was identified inside the field equipment which activates the missiles and it would have made the system inoperable. The threat to Britain's defenses posed by the computer bug was outlined by George Robertson, the Defense Secretary. [Well, at least the failure mode was to not fire.]
2,000 Texans get false overdraft notes in Y2K test Reuters, Detroit News, 12/17/1998 Bank One Texas was testing their Y2K systems to see if they could send out overdraft notices after "1/1/00". They were able to print over 2,000 fabricated notices. But someone forgot to throw away the printouts, which were mailed out! <http://www.detnews.com/1998/technology/9812/17/12170189.htm>
Just to note that the Wassenaar Agreement explicitly exempts 'public domain' software, in its 'General Technology Note': Controls do not apply to "technology" "in the public domain", to "basic scientific research" or to the minimum necessary information for patent applications. The 'General Software Note' has this to add: The Lists do not control "software" which is either: 1. Generally available to the public by being: a. Sold from stock at retail selling points without restriction, by means of: 1. Over-the-counter transactions; 2. Mail order transactions; or 3. Telephone call transactions; and b. Designed for installation by the user without further substantial support by the supplier; or N.B. Entry 1 of the General Software Note does not release "software" controlled by Category 5 Part 2. 2. "In the public domain". See <URL:http://www.wassenaar.org/List/GTNGSN.doc>. Category 5 is the section on 'Information Security', of course. Last updated 10th December 1998. "In the public domain" is defined as: This means "technology" or "software" which has been made available without restrictions upon its further dissemination. N.B. Copyright restrictions do not remove "technology" or "software" from being "in the public domain". See <URL:http://www.wassenaar.org/List/Def.doc>. Martin
The discussion in RISKS-20.10 about theft of infrared codes for door locks brings to mind two similar issues. 1. Conventional RF garage door openers left in cars at parking/repair garages are easily opened so "the baddies" can see the code and program a "universal" opener to get in — many folks don't lock the door between their garage and house. 2. My household internal alarm system uses coded infrared beams. These are trivially defeated with one of those X-10 type "control extenders". A device (you may have seen them, they're 3" pyramid shaped things) that you zap your IR remote control at, this device converts the signal to RF that goes through walls and the receiver reconverts and emits IR to the sensor. (I use one to control my cable box from the bedroom.) Using this thing I can just put the pair between the alarm transmitter and receiver and walk on through. ...wex
I have heard that one can get stand-alone "diagnostic" boxes to record and play back transmissions for the RF car interfaces. In fact, in response to this some vehicles with RF interfaces already have cryptographic encoding of transmissions (I designed such a system with Alan Finn back in my industrial research lab days, and it is in moderately wide use in some US vehicles today). One of the RISKS of this area is that in general the market won't currently bear the cost of high-strength cryptography — they only get a few cents extra to spend on it, and even something like DES tends to cost too much. So it is difficult for consumers to evaluate whether they're getting crypto that is strong enough to be reasonable for the application, or one of the systems that says it is but really isn't (or one of the systems that really is but competitors claim it isn't, or ...). Phil Koopman — firstname.lastname@example.org — http://www.ece.cmu.edu/~koopman
I received a postcard this weekend from the Internal Revenue Service, notifying me that my wife and I have been selected to (optionally) participate in a pilot implementation of the new, improved "e-File". This is a way of electronically filing our income tax returns next April, where not only is the information transmitted electronically and the payment or refund handled via EFT, but we will not even be required to mail in a signature card. Instead of the physical signature card, we can simply use the electronic "signature" number enclosed in a sealed section of the postcard that we received this weekend. (The number is only five digits, but for a one-time use that's probably not too bad.... right?) It's an old story, I know, but one which apparently bears repeating, because organizations are still doing it. This piece of mail was unsolicited, so had someone pilfered it from my mailbox and used it to file a false return, *I* would have been the one in trouble with the IRS. - Andrew Greene [Nice palindromic subject line (ignoring punctuation)! PGN]
In RISKS-20.12 David Wittenberg reported on two situations, one where TCAS may have created an incident that was "saved" by air traffic controllers and another where air traffic controllers may have created an incident that was "saved" by TCAS. He asks "How are pilots to know which one to trust when they must make decisions quickly?" I recently completed a case study on TCAS for a graduate seminar on building safe systems. My students and I were convince, after looking at the extensive TCAS safety activities over the last 15 years, that TCAS is indeed a very effective "safety net" for pilots. After TCAS introduction began, it became clear that a clear set of TCAS operational procedures for both pilots and controllers was necessary to get the full safety benefits of TCAS. For instance: - "pilots should follow resolution advisories unless doing so jeopardizes the safe operation of the flight or the flight crew has definitive visual acquisition of the intruder." - "when responding to a resolution advisory hat directs a deviation from an ATC clearance, pilots should communicate with ATC as soon as possible." - "after a controller has been informed that an aircraft is responding to a resolution advisory, the controller should not issue control instructions to that aircraft that are contrary to the resolution advisory." and so on ... I would personally be very upset if pilots, at this stage in the evolution of TCAS did not follow TCAS resolution advisories. The probability of TCAS errors is very small compared to the likelihood that an air traffic control countermand to a TCAS advisory will lead to potential problems. Unfortunately, I have not seen any reports of the FAA's (or anyone else's) investigation into the former (more interesting, from a safety perspective) incident. The good news is that analysis of reported incidents like this has historically continued to lead to TCAS improvements. The bad news is that the FAA's investment banker has not seen it fit to continue funding of the ~TCAS Program at levels sufficient to continue TCAS improvements. Indeed, there are some who have questioned whether the safety analysis of the new TCAS version 7.0 (developed primarily for use in European airspace) has been adequate.
Earlier this year I had occasion to return from Johannesburg to London on business. I was booked business class for this journey. On the flight day the airline was a plane down so I and a colleague were transferred to economy on the next plane. All the paperwork ws completed and entered into computer on check-in. When I returned to the office I asked the staff who deal with travel to verify with our travel agent that the downgrade was credited with a refund to the business. This did not happen. Full fare was charged. It took many months to unsnarl the paper trail and it was not clear who should have informed who about what. The way the legalese on tickets is written maybe it was all our fault for traveling! It took us many man-hours to resolve the situation and probably many business travelers do not bother assuming that "the system" sorts out little details. But when "the system" comprises the IS systems of several travel firms and subcontractors programmed to bill any upgrades or additions exactly it only takes one to fail to pass on refunds to give a "heads you lose" situation. Just like the vending machines — they never dispense free drinks but there are often stickers on them saying, "it owes me 17p" (UK pence).
I find the quote from NW surprising in view of the fact that the BA "Airmiles" accounts can only be used from the WWW with an initial PIN sent by snail mail to the account holder's address. Changing the PIN can only be done online using https: access. Unlike in the US, in Britain there is popular reluctance to adopt universal ID such as by Social Security number or phone number.
As not-particularly-competent (but professionally paranoid) lawyers have gotten more and more heavily involved, the range of "Y2K critical" dates has grown. A story has made the rounds that some software uses "9999" in the date field to indicate special processing; hence, software is now supposed to be tested for proper operation on 9 Sep 99. (Why operation *on that date* should be affected is beyond me. Then again, since a date written with a two-digit year and no separators necessarily has to be stored in at least a 7-digit field — there being at least one month with a month number larger than 9 containing at least one day with a number larger than 9 — it's beyond me why one would expect software to use 9999 as a flag; 9999999 is much more likely, and completely safe.) I recently saw proposed language that would require a vendor to certify proper operation on about 15 dates, starting with 9 Apr 1999 and ending some time in 2002. One could make a rough kind of sense of many of the dates -- e.g., they tested proper leap year handling in 2000 and thereafter — but some of them are real head-scratchers. The 9 Apr 1999 date is one of those. Is it that 4 and 9 look alike, so that 9499 (or is it 4999?) might have been used as a magic flag? :-) I worked with a lawyer doing Y2K requirements for a major corporation, and convinced them not to put *any* explicit date requirements in. Rather, the language is all written in terms of proper operation on any dates, and with any input date data, that will foreseeably arise in normal system operation during the expected lifetime of the system. Not only does this avoid getting into silly "my list of dates is longer than yours, hence you weren't exercising due diligence" games, but it also avoids a genuine *bug* in the way most of these provisions have been written: They focus so closely on dates around the year 2000 that they ignore the possibility of problems with dates further away. For example, a vendor could introduce windowing - continue to store two-digit years but have 50-99 always represent 1950-1999 while 0-49 represent 2000-2049 - to easily comply with most Y2K requirements I've seen. That's fine if the field in question represents order dates, but not so good if it represents dates of birth. — Jerry
A new computer virus that attacked 10 MCI Worldcom networks last week is capable of spreading throughout computer networks and scrambling the documents on those networks as it goes. "We've never seen anything this sophisticated in 10 years of doing this," says Network Associates' general manager of network security. "This is a completely new strain of virus and the first we've seen that propagates itself with no user interaction." The "Remote Explorer" virus runs on Microsoft Windows NT servers and affects common programs like Microsoft Word. Users clicking on their Word icon might experience a slight delay, but otherwise would be unable to detect the presence of the virus; meanwhile, the virus is busy corrupting files and spreading to other programs. Microsoft officials say they're "aware of other viruses that have the same characteristics," and Network Associates says it has developed a Remote Explorer detector and is working on a solution to decode the affected files. (*Wall Street Journal*, 22 Dec 1998, Edupage, 22 December 1998) [Sounds like a worm to me. PGN]
I came across the following interesting behavior in Windows 95: if you use the "Date/Time Properties" dialog box to change the month or day of month, the system clock is actually set to that date, without the user hitting "OK" or "Apply." The risk is that the user is changing system properties without really being aware of it--if the user hasn't pressed "Apply" he or she will figure that the change hasn't occurred yet. Although pressing "Cancel" will restore the date, any applications that check the system clock in the meantime will read the altered system time. As a simple example, consider a mail reader that checks for new mail every five minutes by comparing system times. If it checks mail during an interval when the clock is set forward a day and then reset, the mail reader will not check mail again for 24 hours. I don't know how many applications expect and rely on the system time to be monotonically increasing. I've been told that other Windows dialogs, besides the date/time, also prematurely change system settings in similar ways, but I'm not sure what implications other settings may have. I tested this on an NT machine and on a Windows 98 machine that was said to have the recent Y2K patch installed, and noticed the same behavior. Dan Weber, Sandstorm Enterprises, Inc. http://www.sandstorm.net/ 1-617-547-0011 email@example.com
I am running Win98 with a Microsoft module that automatically notifies me of "critical updates" and takes me to a Microsoft Updates web site. The first item on the list of updates was one that was intended to make Win98 Y2K compliant. This module was roughly 1.5MB. Further down the list was a beta IE virtual machine (presumably designed to comply with the Sun Java decision.) This file was roughly 4MB; I decided to pass on it. Having specified what I wanted to download — and install (it is both or nothing) -- I noticed that the download was going slowly. And then I discovered the reason: the file that I was downloading was over 5MB ... and included the Virtual Machine Beta that I had not checked! I cancelled the download and went back to check whether I had made and error. I had not. I repeated the exercise with the same result. My conclusion is that in an effort to show 'good faith' compliance with the Sun Java court order, Microsoft is installing the revised Java engine on the computers of users who have decided against using it. The cheese that they are using to bait their trap is the promise of Y2K compliance. Does anyone who is using Win 98 and the IE 5.0 beta know how to get only the Y2K update?
Recently, we at 2600 Magazine published an article with a few Unix scripts in it. A neat feature of XPress is that it allows you to put XPress specific formatting using plain ascii text, i.e. <i> for italics, <b> for bold and so on. So what do you do if you need to you the > or < symbols (in a script for instance)? Easy, XPress just looks for 2 > or < in a row. So when you print something that says: echo "root2:x:0:1:Root:/:/sbin/sh" <> /etc/passwd it gets truncated to echo "root2:x:0:1:Root:/:/sbin/sh" > /etc/passwd In unix, <> appends something to a file, and > replaces the file, so any command in a script to append something to the end of a file, would actually ERASE the file and REPLACE it with the one thing that was supposed to be added. Eeek!
In the process of using the Netscape 4.5 automatic update feature I was asked to join the Netscape Netcenter. The form requests that you supply an account name, password, name, electronic as well as physical address (to be fair, you are given several options on how this information can be used). The unusual aspect of this form, was an option to include a password hint with the following explanation : "If you forget your password, Netcenter will present you this password hint to help jog your memory. Example hint: 'same password as my bank acct.' " The risks are clear : potential access to your name, address and bank PIN ! Alexander V. Konstantinou [Various comments received on this one. TNX. PGN]
It seems to me that more often than not, the reason a Risk rears its ugly head is because either somebody wrote a piece of software without thinking about what possible states exist for various variables, or (a pet peeve of mine), somebody issued an incomplete error message, or both. [My definition of a "complete" error message is one that tells me not only what is wrong, in a way I can be expected to understand, but also gives me some clue what to do to fix it or where to go to get help.] While the example below is a trivial Risk (I hope), it illustrates both of the above quite well. I just installed my copy of the 1999 version of a widely used personal finances software package, and migrated data from a much older version without incident. The new version includes a new feature (new to me anyway) to show alerts if my charge card balances are "near my credit limit." I was very surprised, on my first entry, to see an alert telling me that one of my cards was near my credit limit. I was even more confused when I examined the account: my balance owed was $0! I eventually figured out their default algorithm for those alerts, by examining several other accounts. They would take the account's credit limit, subtract $3,000, and set the resulting value as the default upper limit on the balance owed before an alert would be issued. That made the default on my gasoline company card with its $700 credit limit be negative $2,300. [Feel free to draw your own conclusions about the person who came up with _that_ algorithm.] When I found the place where you can change those defaults, the offending account wasn't listed, so I can't fix the -2,300 value. I suspect that perhaps someone used a negative number in that field to decide that this account isn't a credit card account and shouldn't be in the list. I appear to be stuck with this incorrect warning message until such time as the gas company owes me more than $2,300 on the account. As for that (scary) error message, "Account XXX is near your credit limit", it could have almost trivially been presented to me as the following: "You owe $0 on Account XXX, which has a $700 limit. You will see this alert whenever your balance exceeds $-2,300." The first message wasted a considerable amount of my time in having to figure out what was going on. The second would have made it a lot easier, if not obvious, and would certainly have generated less stress. Millions of people use this program- millions of minutes can get wasted when people can't figure out what is triggering a warning. Is that a trivial Risk? Maybe not. Can I conclude anything else about this incident? Yes. The QA people who tested this, who apparently don't have any charge accounts with credit limits under $3,000, are quite likely overpaid. :-) --Flint Pellett
The National Information Systems Security Conference (NISSC) welcomes papers, panels, and tutorials on all topics related to information systems security. Our audience represents a broad range of information security interests spanning government, industry, commercial, and academic communities. Papers and panel discussions typically cover: * research and development for secure products and systems, presenting the latest thinking and directions; * electronic commerce; * legal issues such as privacy, ethics, investigations, and enforcement; * practical solutions for government, business and industry information security concerns; * network security issues and solutions; * management activities to promote security in IT systems including security planning, risk management, and awareness and training; * implementation, accreditation, and operation of secure systems in a government, business, or industry environment; * international harmonization of security criteria and evaluation; * evaluation of products, systems and solutions against trust criteria; * tutorials on security basics and advanced issues; * security issues dealing with rapidly changing information technologies; * highlights from other security forums; and * implementing policy direction. For more details see http://csrc.nist.gov/nissc/call.htm. [The most important detail: the deadline for submissions is 15 Feb 1999. PGN]
Please report problems with the web pages to the maintainer