Obviously, there's still work to be done, but that's why there are numbers higher than 1.0, right? :-) -matt URL: http://www.research.megasoft.com/people/cmcurtin/snake-oil-faq.html Version: 1.0 Archive-name: cryptography-faq/snake-oil Posting-Frequency: monthly Snake-Oil Warning Signs Encryption Software to Avoid Copyright © 1996 Matt Curtin $Id: snake-oil-faq.html,v 1.0 1996/09/27 21:15:04 cmcurtin Exp $ Distribution Distribution of this document is unlimited. We're specifically interested in reaching people making decisions about what sorts of crypto to use (if any at all), both for their organizations and for themselves, especially those who are non-experts in the field of cryptography and security. This is a work-in-progress. Feedback is greatly appreciated. The Snake Oil FAQ is posted monthly to cypherpunks, sci.crypt, alt.security, comp.security, comp.answers, and comp.infosystems. Disclaimer All contributors' employers will no doubt disown any statements herein. We're not speaking for anyone but ourselves. This is a compilation things that are common among snake oil vendors. It cannot be the sole metric by which a security product is rated, since there can be exceptions to most (or all?) of these rules. (But if you're looking at something that sounds familiar on several of the 'things to watch out for,' you're probably dealing with snake oil. From time to time, a reputable and decent vendor will produce something that is actually quite good, but will use some braindead marketing technique, so be aware that exceptions to general rules can exist.) Every effort has been made to produce an accurate and useful document, but the information contained herein is completely without warranty. If you find any errors, or wish to otherwise contribute, please contact the document keeper, Matt Curtin <cmcurtin@research.megasoft.com> History With the rise in the number of crypto products becoming available came a rise in the amount of ineffective or outright bogus products. After some discussion about this on the cypherpunks list, Robert Rothenburg <wlkngowl@unix.asb.com> wrote the first iteration of the Snake Oil FAQ. Matt Curtin took the early text and munged it into its current state with the help of the listed contributors (and probably some others whose names have inadvertently missed. Sorry in advance, if this is the case.) Introduction This really isn't much of a "FAQ" in the sense that one generally expects to see them: in a question and answer format. Perhaps it will be rewritten as such in the future, but currently, it is more traditionally-formatted paper that covers many topics that are the subject of frequently asked questions. Good cryptography is an excellent and necessary tool for almost anyone. However, there is a multitude of products around. Many good cryptographic products are available, both commercial (including shareware) and free. However, there are also some extremely bad cryptographic products (known in the field as "Snake Oil"), which not only fail do their job of providing security, but are based on, and add to, the many misconceptions and misunderstandings surrounding cryptography and security. Why "snake oil?" The term is used in many fields to denote something that is sold without consideration of its quality, or its ability to live up to claims made by its vendor. This term originally applied to that sold in traveling medicine shows, where the salesmen would claim their elixer would cure just about any ailment that a potential customer could have. Listening to some of the claims made some by modern day crypto vendors, "snake oil" is a surprisingly apt name. Superficially, it is difficult for someone to distinguish the output of a secure encryption utility from snake oil: both look garbled. The purpose of this document is to present some obvious "red flags" that people unfamiliar with the nuts and bolts of cryptography can use as a guideline for determining whether they're dealing with snake oil or the Real Thing. For a variety of reasons, this document is general in scope and does not mention specific products or algorithms as being "good" or "Snake Oil". When evaluating any product, be sure to understand what your needs are. For data security products, what do you need protected? Do you want an archiver that supports strong encryption? An E-mail client? Something that will encrypt on-line communications? Do you want to encrypt an entire disk or partition, or selectively some files? How secure is "secure enough?" Does the data need to be unreadable by third parties for 5 minutes? One year? 50 years? 100 years? Is the third party someone's kid sister? An individual? A corporation? A government? Beware of products that are designed for a specific task (such as data archiving, for example), and add encryption in as an additional feature. Typically, it's better to use an encryption utility for encryption, rather than some tool designed for another purpose that adds encryption to its list of features. Some basics The cryptography-faq (found at http://www.cis.ohio-state.edu/hypertext/faq/usenet/cryptography-faq/top.html) is a more general tutorial of cryptography, and should also be consulted. In an effort to make this FAQ more complete, some very basic topics are included below. Conventional vs. Public Key Cryptography There are two basic types of cryptosystems: symmetric (also known as "conventional," sometimes also called "secret key") and asymmetric ("public key.") Symmetric ciphers require both the sender and the recipient to have the same key. That key is applied to encrypt the data by the sender, and again by the recipient to decrypt the data. The problem here is getting the sender and recipient to share the key. Asymmetric ciphers are much more flexible, from a key management perspective. Each user has a pair of keys: a public key and a private key. The public key is shared widely, given to everyone, while the private key is kept secret. If Alice wishes to mail Bob some secrets, she simply gets (and verifies!) Bob's public key, encrypts her message with it, and sends it off to Bob. When Bob gets the message, he uses his private key to decrypt the message. Secrecy vs Integrity: What are you trying to protect? For many users of computer based crypto, preserving the contents of a message is as important as as protecting its secrecy. Damage caused by a modified message can often be worse than that caused by its disclosure. For example, it may be disquieting to discover that a hacker has read the contents of your funds transfer authorization, but it's a disaster for him to change the transfer destination to his own account. Encryption by itself does not protect a message from change. In fact, there are several techniques for changing the contents of an encrypted message without ever figuring out the encryption key. If the integrity of your messages is important, don't just rely on secrecy to protect them. Check the vendor's claims for an explanation of how their product protects the message from undetected modification. The verification of public keys is an important step. Failure to verify Bob's public key leaves open the possibility that Alice is sending her secrets to someone else, who simply claims to be Bob, using a key that has Bob's name on it, but whose associated private key is in the hands of an attacker. Asymmetric ciphers are much slower than their symmetric counterparts. Also, key sizes must be much larger. See the cryptography FAQ for a more detailed discussion of these topics. Key Sizes Some ciphers, while currently secure against most attacks, are not considered viable in the next few years because of relatively small key sizes and increasing processor speeds (making a brute-force attacks - trying every possible key - feasible). The tables below should give some general guidelines for making intelligent decisions about the key length you need. If the key is too short, the system will be easily broken, even if the cipher is a good one. Having stated the above, it is important to note that a common feature of snake oil is to have large keys. Often, the claimed key lengths are much longer than what is practical, usually due to the vendor's confusion between symmetric and asymmetric cipher key length requirements. (For example, a vendor who claims to use a strong symmetric cipher with a 2048 bit key is probably lacking some basic understanding of key length requirements, and requisite computing power for performing various functions with the keys in question.) In [1] and [2], we're presented with some guidelines for deciding appropriate key length. (It is important to note that this is based on the ability to predict computing power 40, 65, and 100 years from now. Major breakthroughs in computing power 30 years from now might render everything on this chart kiddieplay. This is included so the reader will be able to get a reasonable idea of symmetric key length requirements, and have some sort of a guideline for determining whether the key length of the product he's interested in even makes sense.) The following chart appears in [1]. Security Requirements for Different Information Type of Traffic Lifetime Minimum [Symmetric] Key Length Tactical military information minutes/hours 56-64 bits Product announcements, mergers, interest rates days/weeks 64 bits Long-term business plans years 64 bits Trade secrets (e.g., recipe for Coca-Cola) decades 112 bits H-bomb secrets >40 years 128 bits Identities of spies >50 years 128 bits Personal affairs >50 years 128 bits Diplomatic embarrassments >65 years at least 128 bits U.S. Census data 100 years at least 128 bits As mentioned earlier, asymmetric ciphers require significantly longer keys to provide the same level of security as their symmetric cipher counterparts. Here is a comparison table, again, from [1]. (Due to differences between symmetric and asymmetric algorithms, key length comparisons between the two is difficult. The following is intended to give the reader just a general idea of what is roughly comparable, in order to be able to weed out claims of security of, for example, ciphers with 100-bit asymmetric keys.) Symmetric and Public-Key Lengths With Similar Resistance to Brute-Force Attacks* Symmetric Key Length Public-key Key Length 56 bits 384 bits 64 bits 512 bits 80 bits 768 bits 112 bits 1792 bits 128 bits 2304 bits *These key sizes are for public key cryptosystems based on the problem of factoring large integers, and apply to a number of ciphers based on the discrete log problem (difficulty of taking logarithms in a finite field.) A variation of the discrete log problem (known as Elliptic Curve Discrete Logarithm Problem), where the cryptosystem is based on computations on points of an elliptic curve over a finite field, for example, has been shown to be resistant to brute-force attacks with much smaller keys than other discrete log problem-based ciphers. Ciphers based different problems have different key size requirements. Each type of algorithm's key size requirements depend on the mathematical problem on which the system is based. So, it's important to find out what algorithm (or at least mathematical problem the algorithm uses) and key size is used. One without the other is meaningless. Implementation Environment Other factors that can influence the relative security of a product are related to its environment. For example, in software-based encryption packages, is there any plaintext that's written to disk (perhaps in temporary files)? What about operating systems that have the ability to swap processes out of memory on to disk? When something to be encrypted has its plaintext counterpart deleted, is the extent of its deletion a standard removal of its name from the directory contents, or has it been written over? If it's been written over, how well has it been written over? Is that level of security an issue for you? Are you storing cryptographic keys on a multi-user machine? The likelihood of having your keys illicitly accessed is much higher, if so. It's important to consider such things when trying to decide how secure something you implement is (or isn't) going to be. Some Common Snake-Oil Warning Signs The following are some of the "red flags" one should watch for when examining an encryption product * Technobabble The vendor's description of the product may contain a lot of hard-to-follow use of technical terms to describe how the product works. If this appears to be confusing nonsense, it may very well be (even to someone familiar with the terminology). Technobabble is a good means of confusing a potential user and masking the fact that the vendor doesn't understand anything either. A sign of technobabble is a description which drops a lot of technical terms for how the system works without actually explaining how it works. Often specifically coined terms are used to describe the scheme which are not found in literature about cryptology. Further, if the marketing material isn't clear, what reason is there to believe that the instructions are any better? Even the greatest of products, if not used properly, can be rendered useless. If you can't understand what a vendor is saying, you're most likely better off finding something that makes more sense. * New Type of Cryptography? Beware of any vendor who claims to have invented a "new type of cryptography" or a "revolutionary breakthrough". Truly "new breakthroughs" are likely to show up in the research literature, and professionals in the field are typically won't trust them until after years of analysis, by which time they are not so new anymore. Avoid software which claims to use 'new paradigms' of computing such as cellular automata, neural nets, genetic algorithms, chaos theory, etc. Just because software uses a different method of computation doesn't make it more secure. (As a matter of fact, these techniques are the subject of ongoing cryptographic research and nobody has published successful results based on their use yet.) Anything whose authors claim to have invented a new public key cryptosystem without publishing the details or underlying mathematical principles is highly suspect. Modern cryptography is grounded in mathematical theory. The security is based on problems that are known (or widely believed) to be hard to solve. It's important to understand the difference between a new algorithm or cipher and a new product. Engaging in the practice of developing ciphers and cryptographic products is a fine thing to do. However, to do both, at the same time, is foolish. Many snake oil vendors brag about how they do this, despite the lack of wisdom in such activity. The strength of any encryption scheme is only proven by the test of time. New crypto is like new pharmaceuticals, not new cars. In some ways, though, it's worse: if some pharmaceutical company has some bogus stuff out there, people will start getting really sick. If you're using bogus crypto, you likely won't have any idea that your secrets aren't as secret as you think. * Secret Algorithms Avoid software which uses secret algorithms. Security through obscurity is not considered a safe means of protecting your data. If the vendor does not feel confident that the method used can withstand years of scrutiny by the academic and professional crypto community, then you should be wary of trusting it. (Note that a vendor who specializes in cryptography may have a proprietary algorithm which they'll show to others if they sign a non-disclosure agreement. If the vendor is well-reputed in the field, this can be an exception. On the other hand, if you don't know which vendors are and aren't reputable, you can't take their words for it. You're typically best off avoiding that which is secret.) Beware of specially modified versions of well-known algorithms. This may intentionally or unintentionally weaken the cipher. The use of a trusted algorithm, with technical notes explaining the implementation (or better yet, availability of the source code for the product itself) are signs that a vendor is confident about their product's security. You can take the implementation apart and test it yourself. A lock where attackers can see the internal mechanisms, and still not be able to break it is a strong lock, indeed. A common excuse for not disclosing how a program works is that "hackers might try to crack the program's security." While this may be a valid concern, it should be noted that such 'hackers' can reverse engineer the program to see how it works anyway. If the program is implemented properly and the algorithm is secure, this is not a problem. (If a hypothetical 'hacker' was able to get access you your system, access to encrypted data might be the least of your problems.) * Experienced Security Experts and Rave Reviews Beware of any product claiming that "experienced security experts" have analyzed it, but it won't say who (especially if the scheme has not been published in a reputable journal). Don't rely on reviews in newspapers, magazines or television shows, since they generally don't have cryptologists (celebrity hackers who know about telephone systems don't count) to take the software apart for them. Just because the vendor is a well known company or the algorithm is patented doesn't make it secure either. * Unbreakability Some vendors will claim their software is "unbreakable". This is marketing hype, and a common sign of snake-oil. Avoid any vendor that makes unrealistic claims. (If it sounds too good to be true, it probably is.) No algorithm is unbreakable. Even the best algorithms are breakable using "brute force" (trying every possible key), but if the key size is large enough, this is impractical even with vast amounts of computing power. One-time pads are unbreakable, but they must be implemented perfectly, which is, at best, very difficult. See the next section for a more detailed discussion. Some companies that claim "unbreakability" actually have serious reasons for saying so. Unfortunately, these reasons will generally turn out to depend on some narrow definition of what it means to "break" their security. For example, true one time pads are technically "unbreakable" as far as secrecy goes, but only if several difficult and important conditions also hold. Even then, they are trivially vulnerable to known plaintext attacks on the message's integrity. Other systems may be "unbreakable" only until one of the communicating devices (a laptop, for example) is stolen. So, be sure to find out exactly what the "unbreakable" properties of the system are, and decide if the more breakable portions also provide adequate security. Often, less experienced vendor representatives will roll their eyes and say, "Of course it's not unbreakable if you do such-and-such." The point is that the exact nature of "such and such" will vary from one product to another. Pick the one that matches your operational needs the best. * One-Time-Pads A vendor might claim the system uses a one-time-pad (OTP), which is theoretically unbreakable. (Technically, OTP-generated ciphertext has an equal chance of being each possible plaintext. For example, "598v *$ _+~xCtMB0" has equal probabilities of decrypting to "the whole year in", "the hole youre in", and "you are a weenie!") Snake-oil sellers will try to capitalize on the known strength of an OTP. It is important to understand that any variation in the implementation (which is often done to get around the inherent key management problems of OTPs) means that it is not an OTP, and has nowhere near the security of an OTP. An OTP system is not an algorithm. It works by having a "pad" (called such because originally paper pads were used, before general-purpose computers came into being) of random bits in the possession of both the sender and recipient, but absolutely no one else. (The pad must be sent from one to the other securely, such as in a locked briefcase handcuffed to the carrier, and that sort of thing.) The message is encrypted using the next n bits in the pad as they key, where n is the number of bits in the message. After the bits are used from the pad, they're destroyed, and can never again be used. The bits in the pad must be truly random, generated using a real random source, such as specialized hardware, radioactive decay timings, etc., and not from an algorithm or cipher. Anything else is not a one-time-pad. Further, if the keys (i.e., random bit "pads") are provided by the vendor, the quality of these cannot be verified. How do you know that they aren't sending the same bits (or some trivial mutation thereof) to everyone? Or keeping a copy for themselves? Or selling a copy to your competitors or enemies? OTPs are highly impractical for general purpose cryptography, since the need for random bits is very high, and key management is so cumbersome. OTPs are only practical for extremely low bandwidth communication channels where two parties have the means to exchange pads through a different method from that of their messages. (It is rumored that a link from Washington, D.C., to Moscow was (is?) encrypted with an OTP.) A lesson from the VENONA project (see NSA's web site) is that OTPs are seriously vulnerable if a pad is ever reused. It does not take the resources of a government agency to crack a reused pad. Therefore, the real limitation to their practical use is the generation and distribution of truly random keys for them. You have to distribute at least one bit of key for every bit of data transmitted, including any encrypted protocol data that's sent. If you reuse your pads you run the risk of compromising all data sent with the reused pad. The vendor might (or might try to) confuse random session keys or initialization vectors with OTPs. * Algorithm or product XXX is insecure Be wary of anything that makes claims that particular algorithms or other products are insecure without backing up those claims (or at least citing references to them). Sometimes attacks are theoretical or impractical (requiring special circumstances or massive computing power running for many years), and it's easy to confuse a layman by mentioning these. These usually involve either trying every possible combination of bits for form keys, and trying every possible key until a solution is found, factoring large numbers, or some other cryptanalysis that's just as computationally intensive as one of these methods. * Keys and Passwords The "key" and the "password" are not the same thing. The "key" generally refers to the actual data used by the cipher, while the "password" refers to the word or phrase the user types in, which the software converts into the key (usually through a process called "hashing" or "key initialization"). The reason this is done is because the characters a user is likely to type in do not cover the full range of possible characters. (Such keys would be more redundant and easier for an attacker to guess.) By hashing a key can be made from an arbitrary password that covers the full range of possible keys. It also allows one to use longer words, or phrases and whole sentences as a "passphrase", which is more secure. If the system limits the size of the key or passphrase to something that seems too low, it probably is. If the actual "password" is the cipher's key (rather than hashing it into a key, as explained above), avoid it. If the vendor confuses the distinctions between bits, bytes and characters when discussing the key, avoid this product. Convenience is nice, but be wary of anything that puts too much emphasis on ease of use, without due consideration to cryptographic strength. Avoid anything that lets anyone with your copy of the software to access files, data, etc. without having to use some sort of key or passphrase. Avoid anything that doesn't let you generate your own keys (ie, the vendor sends you a key in the mail, or it's embedded in the copy of the software you buy). Avoid anything by a vendor who does not seem to understand the difference between public-key (asymmetric) cryptography and secret-key (symmetric) cryptography. * Lost keys and passwords If the vendor (or a third party) claims it can recover lost passwords (without using a key-backup or escrow feature), avoid it: a flaw is obviously present, and used to retrieve the contents of an encrypted message. If there is a key-backup or escrow feature, are you in control of the backup, or does the vendor or someone else hold a copy of the key? (Is someone else able to recover your key as easily as you can?) Remember, you have no security against someone who has your key. * Exportable from the USA If the software is made in the US, can it be exported? If the answer is yes, chances are it's not very strong. Strong cryptography is considered munitions in terms of export from the United States, and requires approval from the State Department. Chances are if the software is exportable, the algorithm is weak or it is crackable (hence it was approved for export). If the vendor is unaware of export restrictions, avoid the software: the vendor is not familiar with the state of the art. (For example, if someone claims that the IDEA cipher is exportable from the US, while most other vendors (or the State Department!) do not make such assertions, they're probably lacking sufficient clue to provide you with strong cryptographic software.) Because of export restrictions, some legitimate (not-Snake Oil) products may have a freely exportable version for outside of the USA, which is different from a separate US/Canada-only distribution. (Of course, a freely exportable version isn't secure, since it probably just uses a much smaller key, one that could be easily broken.) Also note that just because software has made it outside of the US does not mean that it is exportable: sometimes a utility will be illegally exported and posted on an overseas site. There are no restrictions on importing crypto products into the US, so a foreign vendor can legally offer a single, secure version of a product for the entire world. * "Military Grade" Encryption Many crypto vendors claim their solution is "military grade." This is a term with no real meaning, since there isn't a real metric by which something can be judged "military grade," except for it to be actually used by various armed forces. Since they don't reveal what they're using, it's neither possible to prove nor to disprove something as being "military grade." Some good crypto products unfortunately also use this term. (Watch for this one especially in combination with other snake oil indicators, i.e., "our military grade encryption system is exportable from the US!") Other Considerations Interface isn't everything: user-friendliness is an important factor, but if the product isn't secure then you're better off with something that is secure (if not as easy to use). No product is secure if it's not used properly. You can be the weakest link in the chain if you use a product carelessly. Do not trust any product to be foolproof, and be wary of any product that claims it is. Glossary algorithm A procedure or mathematical formula. Cryptographic algorithms convert plaintext to and from ciphertext. cipher Synonym for "cryptographic algorithm" cryptanalysis To solve or "break" a cryptosystem. escrow A third party able to decrypt messages sent from one person to another. Although this term is often used in connection with the US Government's "Clipper" proposals, it isn't limited to government-mandated ability to access encrypted information at will. Some corporations might wish to have their employees use cryptosystems with escrow features when conducting the company's business, so the information can be retrieved should the employee be unable to unlock it himself later, (if he were to forget his passphrase, suddenly quit, get run over by a bus, etc.) Or, someone might wish his spouse or lawyer to be able to recover encrypted data, etc., in which case he could use a cryptosystem with an escrow feature. initialization One of the problems with encrypting such things as files vector in specific formats (i.e., that of a word processor, email, etc.) is that there is a high degree of predictability about the first bytes of the message. This could be used to break the encrypted message easier than by brute force. In ciphers where one block of data is used to influence the ciphertext of the next (such as CBC), a random block of data is encrypted and used as the first block of the encrypted message, resulting in a less predictable ciphertext message. This random block is known as the initialization vector. The decryption process also performs the function of removing the first block, resulting in the original plaintext. ITAR International Traffic in Arms Regulations. These are the rules by which munitions (including cryptography), as defined by the US State Department, may (or may not) be exported from the US. key A piece of data that, when fed to an algorithm along with ciphertext, will yield plaintext. (Or, when fed to an algorithm along with plaintext, will yield ciphertext. random session This is a temporary key that is generated specifically for key one message. Typically, in public key cryptosystems, the message to be sent is encrypted with a symmetric key that was specifically generated for that message. The encrypted version of that message, as well as the associated session key can then be encrypted with the recipient's public key. When the recipient decrypts the message, then, the system will actually decrypt the message it gets (which is the ciphertext message and the symmetric key to decrypt it), and then use the symmetric key to decrypt the ciphertext. The result is the plaintext message. This is often done because of the tremendous difference in the speed of symmetric vs. asymmetric ciphers. Contributors The following folks have contributed to this FAQ. Jeremey Barrett <jeremey@forequest.com> Gary Ellison <gary.f.ellison@att.com> <fifersl@ibm.net> Larry Kilgallen <KILGALLEN@Eisner.DECUS.Org> Dutra Lacerda <dutra.lacerda@mail.telepac.pt> <geeman@best.com> Jim Ray <liberty@gate.net> Terry Ritter <ritter@io.com> Robert Rothenburg <wlkngowl@unix.asb.com> Adam Shostack <adam@homeport.org> Rick Smith <smith@sctc.com> Randall Williams <ac387@yfn.ysu.edu> Jim Ray <liberty@gate.net> References 1. B. Schneier, Applied Cryptography, second edition, John Wiley & Sons, 1996 2. M. Blaze, W. Diffie, R. L. Rivest, B. Schneier, T. Shimomura, E. Thompson, M. Wiener, "Minimal Key Lengths for Symmetric Ciphers to Provide Adequate Commercial Security," available from ftp://ftp.research.att.com/dist/mab/keylength.ps -- C Matthew Curtin MEGASOFT, INC Chief Scientist I speak only for myself. Don't whine to anyone but me about anything I say. Hacker Security Firewall Crypto PGP Privacy Unix Perl Java Internet Intranet cmcurtin@research.megasoft.com http://research.megasoft.com/people/cmcurtin/