scans the barcode, the application uses the derived identifier to lookup the current price. In addition, the backend also provides discount information for qualifying products. The backend also decreases the number of available products of that kind
and notifies the manager if the amount falls below a certain threshold.
Figure 2A simplified RFID system
This section describes how RFID tags work in general, what types of tags exist and how they differ. The three frequency ranges that RFID tags typically use are LF, HF, and UHF. Also the difference between passive, semi-passive, and active tags was explained and their advantages and disadvantages were compared. The section concluded by looking at different standards and showed the great interest of the industry by counting the number of issued and backlogged patents .
4. Security
The expected proliferation of RFID tags into the billions has raised many privacy and security concerns. A common concern is the loss of privacy when companies scan tags to acquire information about customers and then using data mining techniques to create individual profiles. This section describes possible scenarios where RFID tags can be exploited. Then it describes what mechanisms exist to defeat those threats or at least make them harder to execute. After that the section concentrates on attacks that are directed against RFID systems.
As RFID technology becomes more sophisticated and item level tagging promises more control and large savings in the supply chain management, companies are tagging items within their production process. To maximize the benefits companies start to require their suppliers to label all items delivered to the company.
5
For example, Wal-Mart, Proctor & Gamble, and the US Department of Defense require their suppliers to phase in item-level tagging. However, products are not the only entity tagged. Animal tagging is quite common at large farms to keep track of their moving \Also, tagging of humans started to appear. In the Spanish Baja Beach Club, VIP members can get an implant that they can use to pay for their drinks in the club. The implanted tag is a VeriChip.
Anti-RFID activists created a few scenarios to show possible exploits if no precautions are taken. The most common one the unauthorized scanning of tags in order to create user profiles. Other scenarios are scanning the medication a person is carrying to conjecture what illness the person might suffer, or a mugger scanning a crowd of people and singling out a person carrying many valuable items (even money, if tagged as proposed). If tags are replacing credit cards eavesdropping becomes also a problem and must be addressed. The above mentioned issues are privacy concerns, but they are not the only issue. Authentication is also needed. For example, newer tags have rewritable memory available to store extra information during the production process. If stores rely on that information to determine the sales price for example, care must be taken so that customers do not change the type of the item to a cheaper one using portable readers. Also the kill command, a mechanism to permanently disable a tag, must be protected from unauthorized access. Recently a paper raised some concerns in the RFID community that claimed that cell phones can be reprogrammed to disable HF tags. In case that tags carry personal information (such as medical history, credit card numbers) a reader has to be authenticated and authorized before it is allowed access the data. In the previous examples the reader has to authenticate to the tag, there are also scenarios when the tag has to authenticate to the reader, for example to detect forged tags. 4.1 Privacy
This section describes methods to ensure privacy. This are usually mechanisms that kick in after the customer bought the product. They are either enabled at time of purchase or controlled by a user-owned device.
Kill command: A command supported by the EPC Class 1 and 2 tags. The command will render the tag unusable once received. To prevent an adversary to call those commands they are password protected, EPC Class 1 tags have 8 bit passwords and EPC Class 2 tags have 32 bit passwords. A theoretical paper described how to reprogram a cell phone with a firmware update to make it scan for tags and once found to quickly enumerate over all possible passwords. A more intelligent approach
6
is described in where the password for the kill command can be discovered by using power analysis on the back scattered signal. The power analysis works since the strength of the back scattered signal depends on how much power the chip on the tag drains which in turn depends on the amount and type of computation it does. Newer chips try to design the circuitry so that power analysis is not possible anymore.
Sleep command: A tag cannot always be killed. Killing a tag on a library book would require retagging the book upon return and therefore defeating the purpose. And yet the privacy of library uses should be protected. The sleep command works similar to the kill command. Once received the only command accepted is the password protected wakeup command. The sleep command suffers from the same problems as the kill command.
Relabeling:describes an approach where the customer can relabel the item tag with a string of user's choice. Some of the old information however remains in a password protected area. The idea is to make that protected information available when the product is discarded so that recycling plants can easily sort items by material.
Split approach: For this approach the information is distributed over two tags and one of the tags is removable by the customer (for example a paper tag on clothing). The fixed tag stores just general information such as the type, care information, etc. of the product while the removable tag contains the serial number. This approach allows item tracking by its unique identifier and still allows the customer to keep track of its own items.
Proxy approach:describes the RFID Guardian. It assumes that all tags are protected by a PIN that the user can set. Once an item is bought the guardian sets a new PIN. If another reader wants to have access to the data stored on a tag the reader requests that information from the guardian which retrieves it from the tag and forwards it to the reader if the reader is authorized.
Distance approach: As described in, tags use the signal-to-noise ratio to get a rough estimate of the readers distance. The closer the reader, the more information is released. For example, scanning an item from far away returns general information such as \a certain brand\and close range scanning reveals finally the serial number. The advantage of the scheme is that it does not require customer actions while still providing those benefits. However, those tags are likely to be more expensive.
7
Blocking approach: A rather crude approach is the following: A special tag that does not follow the medium access protocol is used. RFID tags use a special protocol that controls the access to the shared medium (air). When a reader is in an area with multiple tags it first discovers all tags in its range and then it polls each of the tags. The special tag suppresses that mechanism by back scattering a random signal, practically jamming the frequency used. The example given in the paper is similar to the following. Items are bought in a supermarket, scanned at the checkout point and then placed in a plastic bag with the blocker tag. While the bag is carried home nobody can scan the content of the bag. At home the items are removed from the bag and placed in the fridge. The fridge can then scan the items and add it to the inventory. Instead of implementing that functionality in a tag it can also be implemented into a cell phone for example which creates a safe bubble around its carrier.
Many more approaches exist to protect the privacy of customers but they cannot be discussed here for space reasons. A common problem in general with privacy in case of RFID is that tags are usually small and often embedded, so most people are not aware of them at all. Similar, scanning of tags happens also without people noticing. Some papers proposed the deployment of reader devices that notify their surroundings if an unauthorized reader becomes active. Examples for possible deployment of those devices are hospitals or other controlled facilities where confidential information is often exchanged. 4.2 Authentication
The goal of authentication is to make sure that an entity is what it claims to be. In the context of RFID it means that tags can distinguish authorized readers from other readers. This can be done by using encryption with a preshared key. The other way around is much more difficult. Here a reader has to ensure that the tag it is reading is not altered or copied. As it turns out it is a rather hard problem. Encryption is typically used to establish some trust between both participants of a conversation (in addition to privacy). The main problem for this approach is the very limited resources on the tag itself. Most tags have only a few hundred logic gates, but most encryption schemes require several thousand gates. Several lightweight encryption protocols have been implemented such as AES and. However, it has been shown that they have many weak points and can be broken. For example the Digital Signature Transponder (DST) algorithm protecting theSpeed Pass was broken by researchers from the John Hopkins University allowing them to take gas with a cloned speed-pass.
In addition to weaknesses in encryption algorithms themselves, RFID tags
8
provide unwillingly more help to break those algorithms. Many current tags \lower layer properties, such as the power and timing of the back scattered signal and the processing delay which differ from input to input. That extra information can be used to break encryption even more easily. Newer tags try to fix that problem by two independent circuits for computation and back scattering.
The hope that future generations of RFID tags will provide more resources to implement stronger encryptions might not come true. The reason for this is that there is (and there might always be) the price pressure that demands cheaper tags for item level tagging. And more resources mean higher prices. 4.3 Attack Ranges
In this section I am presenting different ranges that become interesting in terms of security. At the first look, the transmission range specified in the standard seems to be the only range that an intruder is interested in; however, it turns out that ranges far beyond the specified range can be used to gather information about a tag.
The following five ranges have been discussed in and should be kept in mind when designing a new security protocol.
Nominal reading range: This is the range specified by the standard. It is the range in which a sender conforming to the standard can communicate with the tag.
Rogue reading range: This is the range for which a modified sender can communicate with the tag. The modifications can include: sending a signal at higher power than the standard specifies or having a high-gain antenna or antenna array. Needless to say that those modifications can increase the range dramatically.
Tag-to-Reader eavesdropping range: Here the misbehaving reader listens to the back scattered signal when the tag is queried by another reader. Note that the misbehaving reader is passive in this case. This range is larger than the rouge scanning range since the reader does not need to power the tag which is usually the read-range limitation.
Reader-to-Tag eavesdropping range: Here the misbehaving reader listens to the signal sent by the reader that queries the tag. That signal can be read from several kilometers away since the reader has to provide a strong enough signal to power the tag. Note that the misbehaving reader is passive again. This range is larger than the previous one. A misbehaving reader that is able to observe the Tag-to-Reader communication can also observe the Reader-to-Tag communication, therefore getting a full transcript of the communication.
Detection range: This is the range where it is possible to detect the presence of a
9
