Identity Crisis

As a computer professional and IEEE member for 17 years, I was deeply insulted when IEEE Spectrum arrived with “ID Theft costs hundreds of billions a year” emblazoned on its cover. Not only is this false, the article itself [“A Touch of Money,” July] assailed the whacky methodology of the Aberdeen Group. The global trade in drugs is hundreds of billions a year, US $100 billion at wholesale and $300 billion at retail, according to the United Nations Office on Drugs and Crime’s World Drug Report 2005 (Vienna, June 2005), p. 127. When you say more criminals make their money with ID theft than drugs, you make all engineers look like they can’t do math.

The rest of the article is an advertisement for emerging biometric technologies. The authors present no evidence that the ID theft problem is a technical problem. It is a fraud problem. Abraham Abdallah committed bank fraud at Merrill Lynch and wire fraud with the online purchases. If his fraud was made easier by the use of the Internet, then we have to ask who caused fraud to be made easy. Was it Thomas Siebel, whose only action was to have money on deposit at Merrill Lynch? No. It was Merrill Lynch, which somehow deployed a system where $10 million could be transferred without any real authentication.

We need fraud laws more generally like the laws for credit card fraud. Let Merrill Lynch management know they will have to return $9 999 950 of their profits to Siebel unless they can prove he authorized the transfer. The motive for good security will be with the people who field the systems—problem solved.

With motives in the right place, you can post everyone’s Social Security number on the Internet. Banks will pay engineers to deploy the most cost-effective solution. Who knows, they might need to use some new biometric technology. I personally doubt it, but at least then we’d be having an engineering discussion of the costs versus benefits of different technologies. That’s the sort of content I expect in the flagship publication of the engineering community.

Randy Saunders

IEEE Member

Clarksville, Md.

The article makes the proposed system sound better then it really is. First of all, there’s no need to get the fingerprint from a glass, for example. A stolen wallet will have parts of the print on the credit cards since you use it as an authentication method. Although these are not complete, a carefully designed algorithm can reconstruct the complete fingerprint out of the available chunks. Furthermore, the anticipated scheme still uses a magnetic stripe on the card. The stripe is simply waiting for an okay to be released, and a simple override can send a fake okay.

Additionally, the authors did not mention the “Failure to Enroll” problem. In this last case, a lack of unique features or sufficient fingerprint data prevent the person from enrolling into the system. For example, construction workers use their hands for heavy work, which often causes worn out and hidden fingerprints. Finally, using the system for an online transaction does not really change anything other than giving an eavesdropper more data to work around.

Richard Kheir

IEEE Student Member

Villanova, Pa.

The fundamental problem with online shopping nowadays is that it is, technically speaking, an “open-loop” system, which may produce a “steady-state error” in the form of a fraudulent charge on one’s bill. To get rid of this problem, the system should be designed to include the account holder in the “feedback path.”

Consider this scenario. When you apply for a credit card, you specify your e-mail address. When placing an online order, this address will be verified along with your billing information, and a unique order code will be sent to the address. You will have to retrieve this code and use it to complete your online transaction. However the system is designed, it should use the principle known to engineers since the steam engine era: close the loop!

Konstantin Louganski

IEEE Student Member

Blacksburg, Va.

Assessing Metcalfe’s Law

There’s a simple measure of the value of networks that is ignored by all the models discussed in “Metcalfe’s Law Is Wrong” [July]. Each additional connection is only as valuable as the new information that it adds to the body of knowledge accessible via the network. New people joining a network will bring knowledge that is mostly already represented therein, at best adding just a few new items. The probability that a new member will add new knowledge decreases exponentially with the number of people already in the network.

The cost of querying the network, however, as measured by the amount of time spent by the members of the network responding to queries, increases linearly with the number of people in the network. When the incremental cost of a new connection exceeds the information value that it adds, the value of the network starts decreasing. Large networks may therefore actually have less information value to their members than smaller networks.

Donald E. Strebel

IEEE Member

Columbia, Md.

Thank you for the thought-provoking article, “Metcalfe’s Law Is Wrong”. The authors are, however, somewhat harsh in their assessment of Metcalfe’s Law by judging it on its applicability to the market value of network service providers. It is important to clearly define the terms in any formula or law, and in its original form the value was clearly understood as the value derived from connectivity. In this interpretation, the law is intuitive, as connectivity is straightforward to determine in the case of an Ethernet local area network or even the early Internet.

The assumption that the value of each additional connection is arbitrary but constant would indicate that we are talking here about potential value as opposed to actual value. To say that the law must be wrong because we don’t observe the massive merging or interconnecting of networks is to take a far too simplistic view. There are many factors that limit the actual value achievable by network operators, such as the cost and availability of capital, operational costs, technical interoperability, cultural differences, regulation, pricing, and operator politics, to mention a few. For in-depth articles on network economics, the authors may wish to consult Nicolas Economides’ excellent site at http://www.stern.nyu.edu/networks/site.html.

On the technical side, we must take into account the fact that the Internet has become fragmented by software diversity, firewalls, network address translation gateways, and the like. This no longer corresponds to the original end-to-end connectivity model assumed by Metcalfe's Law, which says: “The systemic value of compatibly communicating devices grows as the square of their number.”

I would like to see the original law reinstated and used as a driver for the communications industry—much as Moore’s Law has been used in the semiconductor industry. Its role is to encourage the network operators to maximize the potential value for their users by restoring unencumbered end-to-end connectivity across their networks. The resulting value perceived by the users will ultimately drive the growth of the industry’s actual market value—subject, of course, to basic economic principles.

Dave Penkler

IEEE Member

Grenoble, France

The Differential Hybird

The note from Robert G. Schaffrath [Forum, July] is itself “not quite correct.” The “new” class of hybrids, as defined by the Toyota Synergy Drive Hybrid, is neither a series nor a parallel hybrid. I consider this system a “differential hybrid.”

The differential hybrid configuration uses two motor/generators, a mechanical differential power combiner and a battery with associated controllers to split the output of the internal combustion engine into two channels of power flow. It can assume any state between a full series hybrid and a full parallel hybrid.

In one extreme condition, the power flow is purely mechanical, as in the parallel hybrid. As Schaffrath correctly indicates, this is the highest efficiency mode to deliver the internal-combustion-engine power to the drive wheels. Unfortunately, this does not allow the internal combustion engine to operate at its most efficient point over a range of vehicle conditions.

In the other extreme condition, equivalent to a series hybrid, all the internal combustion engine power is delivered to the drive wheels via the dual motor/generator set. The advantage here is that the internal combustion engine’s operating point can be set to the system optimum by controlling the motor/generator set. The disadvantage is that there are extra losses associated with the motor/generator set.

Obviously, any intermediate power split can be achieved by setting the power split between the two extremes. The cost is the extra motor/generator, control complexity, and the associated losses.

General Motors and Chrysler have recently announced a Dual Mode Hybrid. It is my understanding that it is a differential hybrid with an added transmission in the internal-combustion-engine power-flow path. This approach allows for more of the engine power to be delivered via a more efficient mechanical path. The potential advantage of this system is better highway fuel efficiency.

William C. Follmer

IEEE Life Member

Livonia, Mich.

DMCA: The Darker Side

“Death by DMCA” [June] kindly illustrates some of the impacts of the Digital Millennium Copyright Act. It fails, however, to mention the darker side: the outright theft from the public domain. Current U.S. copyright law has a 95-year term limit, after which works become a part of the public domain. These legally protected digital locks don't simply disappear after the 95-year copyright has expired. In order to legally copy an expired copyrighted work, must one illegally break the digital lock? It is a shame that private interests bring conflict of law into such a broad public forum.

Apparently the authors of the Audio Broadcast Licensing Act and the Analog Hole Bill are also oblivious to these time limitations of copyright law. Adding a copy protection parameter of “ok to copy after date x” would go a long way to complying with current law. And proposals to cryptographically bind legal digital recordings to the initial recording device attempt to take away legal personal usage.

Does anyone actually think that the bound recording device will still be physically functioning when the copyright period expires? Failure to allow transfer to a functioning device takes away fair use once obsolescence or hardware failure set in.

Will those who advocate commercial skipping as theft of advertising time soon seek to make Internet pop-up blockers illegal? I haven’t yet received a check that paid me to watch commercials, so it seems it is my time—not theirs.

Maybe instead of Hollywood spending money to make a mockery of the legal system, it would be better off charging a fair price for its works. Or, if old copyright laws are deemed insufficient for the fast-paced digital millennium, maybe legal protection of the locking mechanisms should be granted—in exchange for a licensing fee, a 20-year copy term, and a free download of the work at the end of the term.

Robert A. Marshall

IEEE Senior Member

Austin, Texas

The implications of the DMCA are indeed outrageous. Hollywood seems to be leaving no stone unturned to kill innovations. Hollywood could demand a royalty from the sale of the gizmos, but putting a curb on the design freedom of the devices is wrong. Why put innovation on the sacrificial altar when it ought to be the wrongdoers who should be punished?

Bharath Bhushan Lohray

IEEE Student Member

Ahmedabad, India

Looking at LIGO

The article “Waiting for Gravity” by Trudy E. Bell [July] gives a clear feeling for the many horrible technical problems facing the designers of the Laser Interferometer Gravitational Wave Observatory (LIGO). But it does not mention one problem that has been bothering me since people started talking about the measurement of gravity waves.

The wavelength of electromagnetic radiation responds to the metric of the space in which it propagates; witness the stretching of primordial gamma rays into the microwave background that we observe today as an effect of the expansion of the universe. If the wavelength responds just like any other length, then the passage of a gravitational wave will stretch and shrink the wavelength of the laser light in the same ratio as the length of the arms of the interferometer, and there will be no phase shift. The result will be no signal, just as in the case of the Michaelson-Morley experiment. Einstein explained that one, but will a similar null effect appear in LIGO? If not, why not?

G. Fonda-Bonardi

IEEE Life Senior Member

Los Angeles

LIGO scientists Frederick J. Raab and Rainer Weiss respond:

The problem is that wavelength is a red herring when considering interferometers. The interferometer measures the differences in phases of the beams returned to the beam splitter, not their wavelengths. The easiest way to think about the phases in an interferometer is as follows:

To avoid problems introduced by worrying over how to connect local and nonlocal observations, imagine the view of an observer sitting on the beam splitter.

To avoid worrying about mixing up things on different time scales, don't worry about the entire cycle of the light wave. Just consider the light that came in from the laser at one point in time (delta function).

Now that you are comfortably seated on the beam splitter, have your assistant, Zweistein, adjust the interferometer mirrors so that the two arm lengths are equal. Here is how you can check on whether Zweistein did it right and gets paid. If the two arm lengths are equal, then the incoming delta function splits into two outgoing lengths (one in each arm).

Sometime later the delta functions return. If no strong gravity waves come by (a good assumption, but you can do many trials if you want to check), then the two delta functions should arrive back at the beam splitter “simultaneously.” Their simultaneous arrival causes them to merge into a single delta function that returns toward the laser, with no light falling onto the photodetector in the so-called “dark” port. In this case, Zweistein should get paid.

Now continue firing delta functions from the laser and wait for a gravity wave to come by. If a gravity wave comes by, space will shrink along one arm and stretch along the other. Since the only thing that is not relative in relativity is the speed of light, the asymmetry introduced into the arm lengths will cause the returning delta functions to arrive at slightly different times, so they cannot merge into a single delta function going back toward the light. This results in light falling onto the photodetector in the “dark” port. Now you can get paid!

Once this gravity wave detection is announced, there will be many arguments in bars and coffeehouses around the world as to whether the detected space-time distortion showed up as a change in space or a change in time, or something in-between. After all, it’s all relative. In the meantime, enjoy the pizza and beer (simultaneously) that you bought after you got paid.

Peter R. Saulson gave a nice description of this result in an article in the American Journal of Physics [Vol. 65, June 1997, pp. 501–505]. The steps are basically the mathematical description of the narrative above. The metric allows you to follow the relation between time and space during the space-time distortion, which changes negligibly over the arms, since they are way shorter than the gravity wave wavelength. The nonrelativity of light speed is given by the expression ds² = 0, where ds² is the squared length of the space-time interval along a light beam. Plug it in and voilà, the delay between the arrival times of the delta functions falls out.

Frederick J. Raab

Director, LIGO-Hanford (Wash.)

This question of the wavelength of the light stretching with the gravitational wave and making it impossible to use light as a means of detecting a gravitational wave has been around from the beginning. It comes from a misunderstanding and also from combining several mutually canceling ideas.

One approach is to let the space not be changed by the gravitational wave but have the gravitational wave exert tidal forces on the masses. This is the approach used by Joe Weber. He found it convenient this way to include the other forces on his acoustic bar detector. In this way of looking at it, there is no change in the way the light behaves and there really is a physical motion of the mirrors. Many of us do not like this approach, because it becomes more difficult to use when in the regions of strong gravitational fields, the regions at the source and unfortunately not at the detector.

The alternative way to think about the process is to have the mirrors stay at fixed locations in the space (like markers on a rubber sheet) and determine the distance between them by timing the light. The wavelength of the light is not relevant to this timing.

Rainer Weiss

Professor emeritus of physics, Massachusetts Institute of Technology, and designer of the LIGO beam tube

Really Robotic

I was interested by Spectral Lines, “Robots Can Ape Us, But Will They Ever Get Real?” [July]. The subject of sentient robots or machines capable of reproducing themselves is a fertile area for science-fiction writers. Since the imaginings of science fiction are often prophetic of actual developments in the future, we should pay close attention to the possibilities they raise.

Of the few science-fiction thinking machines I recall, HAL in 2001: A Space Odyssey is the most benign. While HAL decided to kill off the humans on the mission because they might interfere with its goals as HAL saw them, HAL could not reproduce itself. The visions of Fred Saberhagen in his Berserker series of novels and the Dune series of novels by Frank Herbert, as extended by Brian Herbert and Kevin J. Anderson, definitely represent worst-case scenarios for self-replicating thinking machines.

Saberhagen’s Berserkers actively seek out life and destroy it but are capable of forming temporary alliances with living beings who are willing to betray their kind. The Omnius Evermind of the Synchronized Worlds imagined by the Dune authors is willing to use humans as slaves and vassals but sees little value in keeping them around if they interfere with the plans of the Evermind. Both were creations of people who thought they were doing good but lost control of their charges. If we are to learn anything from history and from the visions of possible futures revealed to us, we must make sure that humans keep control of their creations.

William A. Brewer

IEEE Member

Rochester, N.Y.

Caught in the Ring

I am an “ex-pat” living and working in the Birmingham area, in the United Kingdom, and would like to comment on Justin Mullins’ article “Ring of Steel II” [July]. He states that the public has not voiced an opinion on the privacy issues that go along with 24/7 surveillance. While this may be true of the public spaces video cameras, it is not the case with highway and moving vehicle surveillance. I can attest that the people I work with and am acquainted with absolutely despise the traffic cameras.

The main difference between the acceptance of public space video surveillance and the hostility concerning traffic cameras is simple. Public space cameras tend to catch willful lawbreakers. People who break windows, rob pedestrians, and cause street brawls are recorded, found, and prosecuted.

The traffic cameras catch unintentional lawbreakers. If you are speeding and are caught by a uniformed law enforcement officer, the officer can generally tell the difference if you were willfully speeding or just not paying attention. The officer can then make a value judgment to issue a violation or a warning. In the case of the camera, you are guilty, and any extra circumstances are ignored. That is what gets the average citizen mad. The camera assumes everyone is a suspect.

As people start driving at the speed limit, the net grows tighter. Since overall revenues decrease with compliance, the camera operators must reduce the margin for error and issue violations for as little as 3 miles per hour over the limit of the posted 40 miles per hour. That is probably within the statistical variance of accuracy of both the camera equipment and the passenger car speedometer. I have seen drivers abruptly slow considerably below the speed limit in the camera zones for that reason (only to speed away later). Does that sound like a safe maneuver?

This has less to do with privacy, and more to do with money. The existing London traffic video system is currently being used to charge drivers for the privilege of using the streets of London. The stated reason is an attempt to decrease the traffic flow in the city. However, the revenue brought in by this system tells a different story. Last year the “Congestion Charge” in London brought in about £2 260 000. About half of this was enforcement revenue. Of this, the company in charge of installing, maintaining, and enforcing the system (Capita) gets a cut of the enforcement revenue and a £3 000 000 per month maintenance fee. Transportation for London (a government agency) declared that the system brought £50 000 000 into the city coffers, which leaves the bulk of the money, £176 000 000 , being eaten by the private contractor Capita.

This is looking more like a fishing trip all the time. Londoners are not united on the joys of congestion control. Only an estimated 18 percent reduction in traffic was reported by the Times of London, and that is in a very small area. The Tube is still insanely crowded, and I don’t know what Mullins is referring to when he states that there is “a more pleasant working environment for Londoners.”

As for traffic enforcement using cameras, the story is basically the same. In 2003, there were 4500 speed cameras in use on UK roads, bringing in £1 200 000. During the same time, traffic fatalities in the UK as a whole increased 2 percent and in places like Essex, increased by 24 percent. West Midlands police reported that only 4 percent of motor vehicle collisions were due to excessive speed.

Most of these cameras are placed in order to trap the maximum “fish.” Remember, the stated reason for installation of such devices is to improve law enforcement. Also remember they are paid for by the fines collected. If a camera actually did the job, who would pay for it? With the widespread installation of these camera systems and the claims by the builders of great enforcement gains, the overall police force has stagnated or been reduced.

In the United Kingdom, the police force was decreased from 8900 to 6500 over the time the cameras have been installed. Clearly that has had a detrimental effect on traffic safety. Cameras can only detect speeders. The reckless driver that is involved in or causes the majority of motor vehicle collisions is only blunted by police on the street. In this case, we have placed our trust in the technology when the facts tell us differently. This is another situation where the good intentions of hard-working engineers are turned for profit and away from the designed improvement of life. These statistics are from reports published by the West Midlands Police, The Register (newspaper), and The Guardian (newspaper).

Kenneth Kozol

IEEE Member

Birmingham, England