Hackers Sneak in through the Side Channel
George Lawton
JAN 31, 2013 12:01 PM
A+ A A-

Researchers are looking for answers to a type of emerging cyberthreat that they fear will become a problem with the growth of cloud architectures: the side-channel attack.

Side-channel attacks glean information about a computing or communications system by monitoring changes in its operations as it performs activities such as cryptography or running software. The observed activities could include the electromagnetic signals a computer emits, the amount of time it takes to store data, or the timing of processor operations.

Researchers have demonstrated various types of side-channel attacks, including those against PCs or servers. In the wild, though, hackers have used them primarily against satellite-TV decoders in order to steal programming they didn't pay for.

Security experts are concerned that such attacks might become more common as cloud technology is more widely deployed.

They say that hackers could more easily place their monitoring software on servers that are part of a cloud environment without being detected because such machines routinely host software from many different sources.

Side-channel attacks are unlike typical hacks because they don't require someone to infiltrate or otherwise tamper with a target device, noted security researcher Josh Jaffe, a computer science graduate student at Stanford University.

They are thus difficult to detect. Instead, security requires techniques that prevent attacks to begin with.

A joint Microsoft-MIT research team is developing tools to add subtle programmatic randomness to systems.

This could help prevent some side-channel attacks, which rely on predictable computer processing to let hackers accurately identify differences in a computer's operations under various circumstances.

Microsoft researcher Guy Rothblum said, "While these are preliminary proof-of-concept theoretical results, I am hopeful that in time that can be an important part of a security toolbox for fighting side channel attacks."

The Side Channel

Side-channel attacks on servers would enable hackers to bypass security mechanisms and, for example, read encrypted credit-card numbers. With video decoders, they could identify encryption keys, enabling them to access and copy programming they haven't paid for.

Security experts have known side-channel attacks were possible since the advent of computational encryption.

In the late 1990s, researchers began finding ways to detect computer screens' electromagnetic emissions to determine what was on the monitor and also to hear the sounds of various keystrokes to determine what was being typed.

Now, concern is that the growth of cloud architectures — and their increasing use for sensitive data — will make side-channel attacks more attractive and practical, although properly executing them requires detailed knowledge of a target system.

Unlike direct attacks — which either focus on weaknesses in cryptographic algorithms or use brute computational methods to access a key or secured communication — side-channel attacks use only information observed from the physical implementation of a system.

For example, a malicious program could launch a side-channel attack simply by sending its own data to a target system's memory. From the time the system takes to store and retrieve this data, a hacker could determine how the target system handles information, such as the way it decrypts data.

This process would be complicated, as hackers would have to compile a detailed library that statistically compares the way a machine usually operates with the way it operates when handling various types of data. The library would then have to specify what all of the differences represent.

In some cases, though, this could be incorporated into an analysis tool that would provide the information automatically.

Types of attacks

There are several types of side-channel attacks.

Timing attacks use specially crafted timing loops — software running on the same server as a target application — to measure changes in processing speed and memory usage on a host machine as it runs the loops. Hackers could use an analysis tool to steal information from the computer.

Power-monitoring attacks use oscilloscopes or other devices to monitor the electromagnetic emissions from wires in targeted systems, such as satellite decoders. They could then use an analysis tool to correlate specific emission levels and their timing with a library of commands issued by the target system.

Acoustic cryptanalysis analyzes subtle differences in sounds, such as those from keystrokes, picked up by a stand-alone microphone a hacker places in an office or a compromised computer microphone. This lets hackers identify keystrokes and, therefore, passwords.

Differential fault analysis induces faults — unexpected conditions such as overclocking — into cryptographic systems so that they fail in different ways. To identify encryption keys, cryptanalysts could use a database that correlates the meaning of these failed states with information about cryptographic systems' operations.

Data-remanence attacks use special equipment such as a hard-drive analysis tool to look for magnetic changes in drives and other indicators of sensitive data left behind when a user tried to delete it. Hackers could use a data reader to access some or all of the information.


Side-channel hacks are particularly effective in cryptographic systems such as those used in satellite-TV signal decoders.

This is because hackers have access to the hardware and could steal satellite TV signals they haven't paid for after they crack the encryption.

In fact, noted University of Bristol senior lecturer Elizabeth Oswald, for side-channel hackers, "pay TV is the main attraction. This is why, in terms of effort, the most expensive countermeasures are found for smart cards used in set-top boxes."

One fear is that with the growth of cloud architectures, hackers could install a spy application on a server that hosts a financial application. The application would run specialized timing algorithms on the server to help hackers identify and record sensitive information as applications work with the data.

According to Stanford's Jaffe, hackers could also use the approach to decode the encryption keys used in car locks with wireless remotes, server-based decryption systems, and parking meters that work with smart cards.

Closing the Side Channel

There has been considerable research on ways to counter side-channel attacks.

In the past, researchers have tried to add randomness to secure systems to counter side-channel attacks, but the overhead made them slower. The latest MIT/Microsoft research embeds randomness algorithms into secure code with the least amount of overhead.

These techniques use automated compilers to add random changes in processor-timing patterns for specific routines used in cryptographic and other sensitive computations, explained Microsoft's Rothblum.

The approach reduces overhead by using only the randomness necessary to provide the desired level of security.

In addition, the compiler adds the randomization automatically, whereas previous techniques required time-consuming manual randomization.

The scientists are still working on their technique.

It's becoming harder to architect systems that will be secure at every point in its process, particularly with the rise in consumer media devices with encrypted content and in shared cloud architectures.

Stanford's Jaffe said, "The biggest challenges in addressing side-channel vulnerabilities stem from the fact that [securing] systems can be very complex, and really solving the problem requires [cooperative] effort from the hardware people, software people, and the cryptographic-protocol folks."

by George Lawton

[%= name %]
[%= createDate %]
[%= comment %]
Share this:
Please login to enter a comment: